Methods and systems for query response generation

Abstract

Methods and server systems for query response generation are described herein. Method performed by a server system includes determining, by a set of domain-specific Artificial Intelligence (AI) agents, a activity execution operation to be performed based on an intent associated with a query and query execution information for the intent. Method further includes identifying, by the set of domain-specific AI agents, a set of resolutions for the activity execution operation. Each of the set of resolutions is indicative of an outcome of performing a particular activity. Method further includes determining, by the set of domain-specific AI agents, an insight based on the set of resolutions. Method includes generating, by a master AI agent, a response for the query based on the insight determined by each of the set of domain- specific AI agents. The master AI agent is communicably coupled to the set of domain-specific AI agents.

Description

METHODS AND SYSTEMS FOR QUERY RESPONSE GENERATIONTECHNICAL FIELD

[0001] The present disclosure relates to the field of organization management and, more particularly, to electronic methods and complex processing systems for generating responses to queries for different organizational processes from different users.BACKGROUND

[0002] In a constantly changing world, for an organization to experience potential growth, well-organized management of various organizational processes is required. An organizational process is a set of related, structured activities and steps performed by individuals or equipment within an organization to achieve the organization’s basic objectives, such as increasing profitability and customer satisfaction. Examples of organizational processes include supply chain management, product design, forecasting, quality control, and the like. As may be understood, each organizational process is associated with a set of ad hoc tasks to be performed by different entities for completing the said process or a predefined set of steps associated with a goal. For instance, supply chain management includes tasks, such as procurement, manufacturing, logistics, distribution, etc.

[0003] Generally, various stakeholders are involved in the execution of the tasks that are associated with various organizational processes at different organizational levels. These stakeholders may utilize one or more software tools for the execution of some of the tasks. Some other tasks may need a physical examination and / or manual execution by the corresponding stakeholder. Further, to improve the performance of the organization, the organizational processes are embedded with various advanced features, such as process management, process optimization, process mapping, process simulation, process automation, and the like. In some approaches, these advanced features may utilize advanced technologies such as Artificial Intelligence (Al) and / or Machine Learning (ML) for the execution of organizational processes. Benefits of including such features in the organizational processes include saving time, improved transparency, reduced costs, improved customer satisfaction, and the like.

[0004] Conventionally, Al agents are utilized to perform the operations related to the above-mentioned features. An Al agent refers to an entity that can act upon an environmentP24-016PCT1using sensors and actuators to achieve goals. These Al agents are capable of designing a workflow based on inputs, utilizing available tools, and creating sub-tasks autonomously to generate the output for complex goals. However, Al agents existing in the market are generally designed for a specific purpose. For instance, Al agents can be chatbots and virtual assistants for customer support and services, content-creating agents, surveillance and monitoring robots, and the like. Due to the limitation of being built for a specific purpose, such Al agents are unable to address organizational processes where inputs from various sources have to be considered. Further, such Al agents are unable to perform cross-organizational process-related tasks as well due to their inherent limitations.

[0005] Thus, it is desirable to find technological solutions that improve the performance of an organization by efficiently solving a problem or providing accurate insights for solving the problem faced by a user.SUMMARY

[0006] There exists a need for techniques to overcome one or more limitations stated above, such as the incompatibility of one or more Artificial Intelligence (Al) agents with the purpose of one or more organizational processes associated with an organization. Other limitations include the inability of the Al agents to address the organizational processes where access to various sources is required, inability to perform cross-organizational process-related tasks, impreciseness in the results obtained from the Al agents, incapability to generate precise collective insights from outcomes of the different Al agents, inability to evaluate the efficiency associated with various tasks and sub-tasks that are associated with the organizational processes, and the like.

[0007] Various embodiments of the present disclosure provide methods and systems that enable the generation of tasks and sub-tasks that are more efficient or optimized for managing an organizational process corresponding to an organization. The implementation of the organizational process using the optimized tasks and / or the optimized sub-tasks generates more accurate results. Optimized tasks and optimized sub -tasks also save time, reduce costs, improve transparency, improve customer satisfaction, and the like. Accurate results from each of the organizational processes of the organization lead to a significant improvement in the overall performance of the organization.

[0008] To achieve the above-mentioned and other objectives of the present disclosure, in one aspect, a method for query response generation is disclosed. The method may beP24-016PCT1performed by a server system. The method includes determining, by a set of domain-specific Artificial Intelligence (Al) agents, an activity execution operation to be performed based, at least in part, on an intent associated with a query and query execution information for the intent. The method further includes identifying, by the set of domain-specific Al agents, a set of resolutions for the activity execution operation. Each of the set of resolutions is indicative of an outcome of performing a particular activity associated with the activity execution operation. The method further includes determining, by the set of domain-specific Al agents, an insight based, at least in part, on the set of resolutions. Furthermore, the method includes generating, by a master Al agent, a response for the query based, at least in part, on the insight determined by each of the set of domain-specific Al agents. The master Al agent is communicably coupled to the set of domain-specific Al agents.

[0009] An advantage of some embodiments is that a response to a particular query can be obtained using a master Al agent by receiving insights from different domain-specific Al agents. The response provides an accurate insight for solving a problem presented in the query, intelligently generated by a master-slave operation between the master Al agent and the domain-specific Al agents.

[0010] In an aspect, the method further includes determining, by the master Al agent, the intent associated with the query based, at least in part, on a predefined context of the query and historical intent information of the master Al agent. The step further includes selecting the set of domain-specific Al agents for the intent based, at least in part, on relevancy factors between the set of domain-specific Al agents and the intent. Herein, the master Al agent is configured to operate the set of domain-specific Al agents to generate the insight, based at least on the intent.

[0011] An advantage of some embodiments is that the determination of the intent associated with the query helps the master Al agent select the most appropriate domain-specific Al agent from the set of domain-specific Al agents. This selection helps improve the likelihood of the activities to be performed for generating the insight accurately. With the predefined context of the query, the determination of the intent becomes easy. Also, the knowledge of the intent helps the master Al agent to prioritize and streamline between different activities, optimize resource utilization, maintain consistency, and facilitate complex operations.

[0012] In an aspect, the step of selecting a particular domain-specific Al agent in the set of domain-specific Al agents for the intent includes accessing a plurality of agent profilesP24-016PCT1of a plurality of available domain-specific Al agents. The step further includes calculating a plurality of relevancy factors between the plurality of agent profiles and the intent. Then the step includes selecting the domain-specific Al agent associated with an agent profile that has the highest relevancy factor from the plurality of relevancy factors.

[0013] An advantage of some embodiments is that the determination of the relevancy factor between the agent profiles and the intent of the query helps the master Al agent to select an appropriate domain-specific Al agent. The selection of the domain-specific Al agents based on the relevancy factor ensures a data-driven approach. Such an approach is advantageous as it leads to more accurate, reliable, and unbiased outcomes.

[0014] In an aspect, the step of determining the activity execution operation includes identifying a query category for the intent based, at least in part, on predefined category information, the query category including one of a task, a goal, or a combination thereof.

[0015] An advantage of some embodiments is that the identification of the query category for the intent helps the set of domain-specific Al agents understand whether the query is a task, a goal, or a combination of the task and the goal. As a result, precise activity execution operation can be generated, upon validation and implementation of which, an accurate response for the query can be generated.

[0016] In an aspect, upon identifying the query category to be the task, the step of determining the activity execution operation includes generating, dynamically, a task execution operation including a set of sub-tasks to be performed for generating the insight based, at least in part, on the intent and task-related information. The step further includes assigning the task execution operation to the activity execution operation to be performed

[0017] An advantage of some embodiments is that the identification of the query category to be the task directs the set of domain-specific Al agents to access the task-related information for the determined intent. Upon accessing the task-related information, dynamically generate the task execution operation based on this information which also includes available tool information. As a result, if the task requires one or more sub-tasks to be performed based on the task execution operation, then the set of domain-specific Al agents is designed to determine the set of sub-tasks associated with the task execution operation. Further, the set of domain-specific Al agents, accordingly, access required tools for the implementation of these sub-tasks and generate the set of resolutions.

[0018] In an aspect, upon identifying the query category to be the goal, the step ofP24-016PCT1determining the activity execution operation further includes determining a goal execution operation including a set of tasks to be performed for determining the insight based, at least in part, on the intent and predefined goal execution information. The step further includes assigning the goal execution operation to the activity execution operation to be performed.

[0019] An advantage of some embodiments is that the identification of the query category to be the goal directs the set of domain-specific Al agents to access the predefined goal execution information for the determined intent. The goal can require one or more tasks to be performed which are predefined. In other words, the predefined goal execution information includes a goal execution operation which is predefined. The one or more tasks that need to be performed are based on the goal execution operation. Moreover, using predefined data allows for quicker generation of resolutions, that follows a business / product specific procedure and hence a timely and business / product specific response can be provided for the query. Herein, the business / product specific responses indicate the intent determined to be associated with the query.

[0020] In an aspect, upon identifying the query category to be a combination of the task and the goal, the step of determining the activity execution operation includes segregating the query into at least two sub-queries including a first sub-query indicative of the task and a second sub-query indicative of the goal. The step further includes generating and assigning, dynamically, a task execution operation to the activity execution operation for generating the insight for the first sub-query. Furthermore, the step includes determining and assigning a goal execution operation to the activity execution operation for generating the insight for the second sub -query.

[0021] An advantage of some embodiments is that the identification of the query category to be the combination of the task and the goal directs the set of domain-specific Al agents to segregate the query for the task and the goal. This segregation enables scalability and parallel processing, speeding up response times. This also improves accuracy, as noise is reduced, allowing each of the set of domain-specific Al agents to better understand and response to the specific context and intent.

[0022] In an aspect, the step of generating the insight further includes determining a validity of the activity execution operation based, at least in part, on the activity execution operation, the set of resolutions, and a Large Language Model (LLM).

[0023] An advantage of some embodiments is that the determination of the validity ofP24-016PCT1the activity execution operation using the LLM helps the set of domain-specific Al agents to fine-tune these activities. As a result, an accurate insight into solving the problem stated in the query can be generated.

[0024] In an aspect, the step of determining the validity of the activity execution operation includes accessing the activity execution operation and the set of resolutions. The step further includes determining, by the LLM, a set of fine-tuning parameters for each activity of the activity execution operation based, at least in part, on the intent and a predefined training dataset of the LLM. Furthermore, the step includes transferring, by the LLM, the set of finetuning parameters to the domain-specific Al agent based at least on the intent.

[0025] An advantage of some embodiments is that the determination of the set of finetuning parameters for each activity of the activity execution operation for each of the set of domain-specific Al agents helps to understand inaccuracies associated with the activity execution operation. As a result, the set of domain-specific Al agents will be able to re-generate the activity execution operation that are more accurate based on the set of fine-tuning parameters.

[0026] In an aspect, the step of determining the validity of the activity execution operation further includes receiving, by the domain-specific Al agent, the set of fine-tuning parameters from the LLM. The step further includes identifying the activity execution operation to be valid when the set of fine-tuning parameters matches with fine-tuning criteria. Alternatively, the step includes identifying the activity execution operation to be invalid when the set of fine-tuning parameters deviates from the fine-tuning criteria.

[0027] An advantage of some embodiments is the identification of the validity of the activity execution operation by each of the set of domain-specific Al agents by merely comparing the set of fine-tuning parameters with the fine-tuning criteria. The LLM is fined- tuned based on the intent of the query and preferred set of tasks that need to be performed to generate an accurate insight for solving the problem stated in the query. Thus, the set of domain-specific Al agents can quickly and efficiently fine-tune the activity execution operation using the LLM.

[0028] In an aspect, upon identifying that the activity execution operation is invalid, the method further includes generating a new activity execution operation based, at least in part, on the intent and the set of fine-tuning parameters. The method further includes identifying a new set of resolutions for the new activity execution operation based, at least inP24-016PCT1part, on performing the new activity execution operation. Furthermore, the method includes determining a validity of the new activity execution operation based, at least in part, on the new activity execution operation, the new set of resolutions, and the LLM.

[0029] An advantage of some embodiments is the generation of the new activity execution operation when the activity execution operation is identified to be invalid. This immediate action of the set of domain-specific Al agents to generate this new activity execution operation accelerates the process of generating an accurate insight for solving the problem stated in the query. The validity of this new activity execution operation is checked. It is noted that the continuous examination of the validity of the activity execution operation and the new activity execution operation ensures that there is no compromise in the accuracy of the response that may be generated for the query.

[0030] In an aspect, the step of generating the response includes receiving, by the master Al agent, the insight from each of the set of domain-specific Al agents. The step further includes determining, by the master Al agent, a validity of the insight based, at least in part, on the query and the query execution information. Furthermore, in response to determining that the insight from each of the set of the domain-specific Al agents is valid, the step includes generating, by the master Al agent, the response, based at least on the insight.

[0031] An advantage of some embodiments is that the determination of the validity of the insight generated by each of the set of domain-specific Al agents, at the master Al agent, provides a second-level validation. To that note, if any of the set of domain-specific Al agents provides poor insights to the master Al agent, then this two-fold validation process ensures the generation of an accurate response to the query. In addition, the two-fold validation process is time and resource-efficient as the master Al agent is intelligent enough to re-create required tasks to improve the accuracy of the response for the query.

[0032] In an aspect, in response to determining that the insight from each of the set of domain-specific Al agents is invalid, the method further includes selecting a different domainspecific Al agent for the intent, wherein the different domain-specific Al agent has a second highest relevancy factor.

[0033] An advantage of some embodiments is the capability of the master Al agent to select another domain-specific Al agent if one or more insights received from the selected set of domain-specific Al agents are invalid. This aspect enables the generation of dynamic responses to the queries and maintains the accuracy of the response by re-selecting differentP24-016PCT1domain-specific Al agents if the previously selected domain-specific Al agents do not provide the expected results.

[0034] As per another embodiment of the present 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 communicably coupled to the memory. The processor is configured to execute the instructions to cause the server system, at least in part, to determine, by a set of domain-specific Artificial Intelligence (Al) agents, an activity execution operation to be performed based, at least in part, on an intent associated with a query and query execution information for the intent. The server system is further caused to identify, by the set of domain-specific Al agents, a set of resolutions based, at least in part, on the activity execution operation, each of the set of resolutions being indicative of an outcome of performing a particular activity associated with the activity execution operation. Furthermore, the server system is caused to determine, by the set of domain-specific Al agents, an insight based, at least in part, on the set of resolutions. Moreover, the server system is caused to generate, by a master Al agent, a response for the query based, at least in part, on the insight determined by each of the set of domain-specific Al agents. Herein, the master Al agent is communicably coupled to the set of domain-specific Al agents.

[0035] As per yet another embodiment of the present disclosure, a non-transitory computer-readable storage medium is disclosed. The non-transitory computer-readable storage medium includes computer-executable instructions that, when executed by at least a processor of a server system, cause the server system to perform a method. The method includes determining, by a set of domain-specific Artificial Intelligence (Al) agents, an activity execution operation to be performed based, at least in part, on an intent associated with a query and query execution information for the intent. The method further includes identifying, by the set of domain-specific Al agents, a set of resolutions based, at least in part, on the activity execution operation. Each of the set of resolutions is indicative of an outcome of performing a particular activity associated with the activity execution operation. The method further includes determining, by the set of domain-specific Al agents, an insight based, at least in part, on the set of resolutions. Furthermore, the method includes generating, by a master Al agent, a response for the query based, at least in part, on the insight determined by each of the set of domain-specific Al agents. The master Al agent is communicably coupled to the set of domainspecific Al agents.P24-016PCT1

[0036] The foregoing summary is illustrative only and is not intended to be in any way limiting. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features will become apparent by reference to the drawings and the following detailed description.BRIEF DESCRIPTION OF FIGURES

[0037] For a more complete understanding of example embodiments of the present technology, reference is now made to the following descriptions taken in connection with the accompanying drawings in which:

[0038] FIG. 1 is an example representation of a virtual assistance environment, in accordance with various embodiments of the present disclosure;

[0039] FIG. 2 illustrates a simplified block diagram of a server system, in accordance with an embodiment of the present disclosure;

[0040] FIG. 3 illustrates a flow diagram depicting a set of organizational processes assigned with a set of domain-specific Al agents, in accordance with an embodiment of the present disclosure;

[0041] FIG. 4 illustrates a flow diagram depicting a master-slave system architecture used for determining a solution to a problem through a query response generation process, in accordance with an embodiment of the present disclosure;

[0042] FIG. 5 illustrates a flow diagram depicting a process of evaluating a validity of a activity execution operation generated by the set of domain-specific Al agents, in accordance with an embodiment of the present disclosure; and

[0043] FIG. 6 illustrates a flow diagram of a method for query response generation, in accordance with an embodiment of the present disclosure.

[0044] The drawings referred to in this description are not to be understood as being drawn to scale except if specifically noted, and such drawings are only exemplary in nature.DETAILED DESCRIPTION

[0045] In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. It will be apparent, however, to one skilled in the art that the present disclosure can be practiced without these specific details. Descriptions of well-known components and processing techniques are omitted to not obscure the embodiments herein unnecessarily. The examplesP24-016PCT1used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.

[0046] References in this specification to “one embodiment” or “an embodiment” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. The appearances of the phrase “in an embodiment” in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Moreover, various features are described which may be exhibited by some embodiments and not by others. Similarly, various requirements are described which may be requirements for some embodiments but not for other embodiments.

[0047] Moreover, although the following description contains many specifics for the purposes of illustration, anyone skilled in the art will appreciate that many variations and / or alterations to said details are within the scope of the present disclosure. Similarly, although many of the features of the present disclosure are described in terms of each other, or in conjunction with each other, one skilled in the art will appreciate that many of these features can be provided independently of other features. Accordingly, this description of the present disclosure is set forth without any loss of generality to, and without imposing limitations upon, the present disclosure.

[0048] Conditional language such as, among others, “can”, “could”, “might”, or “may”, unless specifically stated otherwise, are otherwise understood within the context as used in general to convey that certain embodiments include, while other embodiments do not include, certain features, elements and / or steps. Thus, such conditional language is not generally intended to imply that features, elements, and / or steps are in any way required for one or more embodiments or that one or more embodiments necessarily include logic for deciding, with or without user input or prompting, whether these features, elements and / or steps are included or are to be performed in any particular embodiment.

[0049] Disjunctive language such as the phrase “at least one of X, Y, or Z” unless specifically stated otherwise, is otherwise understood with the context as used in general to present that an item, term, etc., may be either X, Y, or Z, or any combination thereof (e.g., X, Y, and / or Z). Thus, such disjunctive language is not generally intended to, and should not,P24-016PCT1imply that certain embodiments require at least one of X, at least one of Y, or at least one of Z to each be present.

[0050] Unless otherwise explicitly stated, articles such as “a” or “an” should generally be interpreted to include one or more described items. Accordingly, phrases such as “a server system configured to” are intended to include one or more recited server systems / processors. Such one or more recited devices can also be collectively configured to carry out the stated recitations. For example, “a processor configured to carry out recitations A, B, and C” can include a first processor configured to carry out recitation A working in conjunction with a second processor configured to carry out recitations B and C. The same holds true for the use of definite articles used to introduce embodiment recitations. In addition, even if a specific number of an introduced embodiment recitation is explicitly recited, those skilled in the art will recognize that such recitation should typically be interpreted to mean at least the recited number (e.g., the bare recitation of “two recitations” without other modifiers, typically means at least two recitations or two or more recitations).

[0051] It will be understood by those within the art that, in general, terms used herein, are generally intended as “open” terms (e.g., the term “including” or “comprising” should be interpreted as “including / comprising but not limited to,” the term “having” should be interpreted as “having at least,” the term “includes” or “comprises” should be interpreted as “includes / comprises but is not limited to,” etc. .

[0052] For expository purposes, the term ‘organization’ refers to an enterprise having a structured group of people who work together to achieve common goals. An organization has a hierarchical structure with defined roles and responsibilities to individuals and operates according to established processes, rules, and policies.

[0053] The term ‘organizational process’ refers to a series of tasks or a set of activities to be performed by a group of stakeholders or equipment within the organization to achieve an organizational goal and / or to solve organizational problems. Organizational processes can be related to different fields or domains associated with operations, finance, marketing, etc., of the organization. Different domains in the operations include inventory management, order processing, customer support, planning, organizing, controlling, product development, and the like. Similarly, different domains in finance include budgeting, accounting, financial analysis, reporting processes, and the like. Further, different domains in marketing include lead generation, negotiation, closing deals, brand positioning, advertising, customer engagement,P24-016PCT1and the like.

[0054] Further, the term ‘activity’ refers to a defined action where information is processed, stored, or transferred as a part of an organizational operation, a business operation, or a physical operation. The activities may be performed by different stakeholders of the organization or equipment within the organization. The activities may be associated with an activity execution operation / plan and are related to an organizational process that is performed to achieve an organizational goal or solve an organizational problem. The activities can be subtasks related to an ad-hoc task that is to be performed to provide a solution to a problem. The activities can be a set of predefined tasks that may be associated with a goal that is predefined for providing a solution to a particular problem. For example, an activity to be performed by a Human Resource (HR) can be a job interview. This activity is related to a process called recruitment which solves the problem of hiring the most appropriate people for open positions in the organization.

[0055] Furthermore, the terms ‘Artificial Intelligence (Al) agent’ or ‘autonomous Al agent’ (used interchangeably herein) refer to a program or an entity that interacts with its surrounding (or environment) by collecting data through sensors and acting through actuators by performing self-determined tasks to meet predetermined goals. Types of Al agents include simple reflex agents, model-based reflex agents, goal-based agents, utility-based agents, and learning agents. It is noted that based on the application and its complexity, a particular type of Al agent may be selected.

[0056] It is understood that for an organization to experience success or improvement in its performance, one or more organizational processes associated with the organization may have to be optimal and efficient. It is noted that these organizational processes are designed to address various organizational problems associated with the organization. The problems surround the management of operations, finance, marketing, compliance and regulatory, quality, etc., of the organization. The implementation of each of the organizational processes requires performing several steps, such as setting goals, identifying inputs and outputs, process mapping, assigning process tasks to different stakeholders or equipment within the organization, performing tests, process implementation, analyzing the results, and the like. However, the implementation of these processes is associated with several drawbacks, such as the setting of unclear goals and objectives, delays, communication breakdowns, duplicated efforts, complex processes, overreliance on manual work, lack of training and support,P24-016PCT1neglecting measurement and monitoring, failing to address resistance to change among the stakeholders and customers, ignoring security risks, and the like. Failing to evaluate the accuracy of these steps, tasks, sub-tasks, and ultimately the organizational processes is one of the reasons for these drawbacks. To overcome these drawbacks, optimization and improving the efficiency of the organizational processes is required.

[0057] Conventionally, various techniques are used for optimizing organizational processes. One approach includes utilizing Al agents existing in the market with respective predefined purposes for organizational processes with matching purposes. In some scenarios, a problem statement is prompted to a particular Al agent as a query. The Al agent may try to understand the purpose or intent associated with the query based on a predefined context associated with the Al agent. Upon understanding the purpose, the Al agent may call for one or more tools in the backend, perform one or more tasks, and generate insight from results obtained from the execution of the tasks. Then, the Al agent provides a solution to the problem as a response to the query based on the generated insight. However, the existing Al agents may not be compatible with the purpose and context of the organizational processes. For example, a manufacturing and warehousing autonomous robot, in general, can take care of assembly lines or warehouses to move items. However, its configuration may not work for the assembly lines and warehouses of a particular organization as it is unaware of the context of this organization. In other words, due to the limitation of being built for a specific purpose, such Al agents are unable to address organizational processes where inputs from various sources have to be considered.

[0058] As a result, it may be required for the organization to build domain-specific Al agents for the domain or context associated with each organizational process. However, domain-specific Al agents are unable to perform cross-organizational process-related tasks as well due to their inherent limitations. Therefore, the overall performance and growth of the organization are still mediocre as human intervention is still required to generate collective insight from different domain-specific Al agents. Also, there is a possibility that the insights generated by the domain-specific Al agents are inaccurate due to inefficiencies associated with activities performed by the corresponding domain-specific Al agents that remain unnoticed.

[0059] To that end, various embodiments of the present disclosure aim to solve the above-mentioned technical problems by providing an approach for generating a response that is an accurate solution to a problem stated in a query. The approach of the present disclosureP24-016PCT1aims to provide actionable insights as a solution to a problem or a response to a query by combining data from multiple systems through multiple Al agents. This is achieved by determining a activity execution operation to be performed by a set of domain-specific Al agents based on an intent associated with the query. Further, upon implementation of the activity execution operation, a set of resolutions is identified. An insight from the set of resolutions may be determined. Finally, a master Al agent generates a response to the query based on a set of insights obtained from the set of domain-specific Al agents, respectively. This helps in understanding the overall system performance or organization performance and making data-driven decisions and / or actions.

[0060] FIG. 1 is an example representation of a virtual assistance environment 100, in accordance with various embodiments of the present disclosure. The environment 100 includes a server system 102, a plurality of first users 104(1), 104(2), and 104(3) (hereafter collectively referred to as ‘first users 104’), a plurality of second users 106(1), 106(2), and 106(3) (hereafter collectively referred to as ‘second users 106’), an organization 108, each coupled to, and in communication with (and / or with access to) a network 110. In a non-limiting example, the server system 102 may be configured to perform one or more operations, such as, but not limited to, determining an activity execution operation, identifying a set of resolutions, determining insights from the set of resolutions, generating a response to a query based on the insights, and the like.

[0061] In various examples, the first users 104 include individuals, such as stakeholders, investors, sponsors, participants, interested parties, employees, responsible individuals, etc., who are responsible for understanding a particular problem to provide a solution within an organization such as the organization 108. The first users 104 are responsible for performing several tasks and sub-tasks using data analytics skills and / or software tools to generate the solution. In a specific organizational process such as a process of providing customer support and services in an organization (e.g., the organization 108), the first users 104 belong to a customer support team. In an embodiment, the first users 104 include customer representatives or human agents whose responsibility is to respond to customer queries and provide solutions to problems faced by them or address casual customer queries.

[0062] In various other examples, the first users 104 include entities, such as chat-bots, self-assist systems, Al agents, a machine interface, or the like interacting with the customers for businesses. In another organizational process such as a process of manufacturing andP24-016PCT1warehousing, the first users 104 include individuals, workers, helpers, autonomous robots, robotic arms, Al agents, etc., that take care of assembly lines or warehouses to move items. In some embodiments, the first users 104 include individuals or entities that are responsible for analyzing solutions that are generated for different problems and generating a collective insight. This collective insight can then be used to determine the overall performance of the organization 108.

[0063] In an embodiment, the second users 106 include customers, clients, organizations, users, buyers, etc., of the products and / or services offered by the organization 108. Further, in one example, the organization 108 includes a company, an institution, a business, an association, a government body, a private body, a non-profitable entity, etc.

[0064] The network 110 may include, without limitation, a Light Fidelity (Li-Fi) network, a Local Area Network (LAN), a Wide Area Network (WAN), a Metropolitan Area Network (MAN), a satellite network, the Internet, a fiber optic network, a coaxial cable network, an Infrared (IR) network, a Radio Frequency (RF) network, a virtual network, and / or another suitable public and / or private network capable of supporting communication among two or more of the parts or components illustrated in FIG. 1, or any combination thereof.

[0065] Various components in the environment 100 may connect to the network 110 in accordance with various wired and wireless communication protocols, such as Transmission Control Protocol / Internet Protocol (TCP / IP), User Datagram Protocol (UDP), 2ndGeneration (2G), 3rdGeneration (3G), 4thGeneration (4G), 5thGeneration (5G) communication protocols, Long Term Evolution (LTE) communication protocols, future communication protocols or any combination thereof. For example, the network 110 may include multiple different networks, such as a private network made accessible by the server system 102 and a public network (e.g., the Internet, etc., through which the server system 102, any of the first users 104, and any of the second users 106 may communicate.

[0066] In one scenario, as the first users 104 are responsible for determining a solution to various problems associated with the organization 108, the first users 104 may use facilities provided by the server system 102. In another scenario, the second users 106 can also access these facilities of the server system 102 upon authentication of their identities and receiving permission from authorities of the organization 108. In a non-limiting scenario, the authorities include administrators, employers, owners, etc., of the organization 108 regulating various responsibilities of the first users 104. In yet another scenario, when the second users 106P24-016PCT1doesn’t have access to the facilities of the server system 102, the second users 106 can ask queries through a website portal of the organization 108. In such a scenario, the first users 104 can use the facilities of the server system 102 and ask for responses to queries received from the second users 106. Upon receiving the responses from the server system 102, the first users 104 may either provide the same responses to the second users 106 or modify the responses accordingly and provide a custom response to the second users 106.

[0067] Each first user of the first users 104 can be associated with an electronic device to interact with existing / potential clients present at different geographical locations, such as the second users 106. Alternatively, each first user may use the electronic device to prompt a query corresponding to an organizational problem to the server system 102. Herein, the query may be prompted with an expectation to receive a response having an appropriate solution to the problem. Similarly, each second user of the second users 106 may also be associated with an electronic device for prompting a query directly to the server system 102 or interacting with the first users 104. The first users 104 and the second users 106 are depicted to be associated with different electronic devices such as desktop computers and mobile devices in the environment 100. However, it should be understood that the first users 104 and the second users 106 may use any electronic device that may be configured to facilitate interaction between two remote individuals (for example, individuals, such as the first user 104(1) and the second user 106(1)). For example, the first user 104(2) may be equipped with a mobile device, such as a smartphone configured to facilitate engagement between the first user 104(2) and the second user 106(2). Examples for the electronic device include, such as, but are not limited to, a mobile device (e.g., a mobile phone, a smartphone, a tablet, etc.), a smart television, a laptop, a desktop, or the like. In the illustrated example, the second user 106(1) is depicted to be associated with a desktop computer, the second user 106(2) is depicted to be associated with a desktop computer as well, and the second user 106(3) is depicted to be associated with a mobile device. Similarly, the first user 104(1) is depicted to be associated with a mobile device, the first user 104(2) is associated with a desktop computer, and the first user 104(3) is also associated with a desktop computer.

[0068] It should be noted that the features of the server system 102 may be accessible to the first user 104(1) through a User Interface (UI) associated with a platform. The platform can be a website, a mobile application, a website application, or the like facilitating the features of the server system 102 to the first user 104(1). The UI may provide a prompt window for the first user 104(1) to enter the query as a prompt. In some embodiments, the platform canP24-016PCT1facilitate the first user 104(1) to provide the query through voice commands, gestures and body language, image or video inputs, sensor data, and behavioral data, by integration with other systems, wearable devices, robotics, automated triggers, or the like. Similarly, in some embodiments, the electronic device associated with the second user 106(1) is also provided with similar UIs or different UIs that facilitate the second user 106(1) to send a query either directly to the server system 102 or to the electronic device of the first user 104(1).

[0069] It should be understood that the second users 106 and the first users 104 may engage in communication with each other and the server system 102 via different applications such as, but not limited to, a mobile application, an email application, a web-based email client application, a web-based chat application, and so on, for a variety of purposes.

[0070] In an example scenario, the first user 104(1) is a manager in the organization 108 responsible for managing the overall performance of the organization 108. If the first user 104(1) is required to know the contact details of a particular customer ‘X’ and the performance of a warehouse UK001 for the month of Feb 2023. This information may be required for the first user 104(1) to make certain decisions impacting the performance of the organization 108. The first user 104(1) may contact the sales team and request the contact details of the customer ‘X’. The sales team may have to search for the name of the customer ‘X’ in a customer dataset of the organization 108. There is a possibility that one or more individuals can have similar names resembling the name of the customer ‘X’. So, the sales team has to search for the exact name of the customer ‘X’ in the customer dataset. Once found, it is sent to the first user 104(1). Simultaneously, the first user 104(1) may have contacted the warehouse team to get information about the performance of the warehouse UK001 for the corresponding month. The warehouse team will perform several tasks to get the performance of the particular warehouse. The tasks can include, for instance, mentioning profitability across all regions, determining a region with the lowest profitability, determining the average profitability of the last three months, comparing the profitability between these months to determine the reason for the performance, and the like. Based on the implementation of these tasks, a final insight into the performance of warehouse UK001 may be determined by the warehouse team and provided to the first user 104(1). Upon receiving the required information from the sales team and the warehouse team, the first user 104(1) may have to analyze and verify this information. Finally, the first user 104(1) can conclude with a collective insight based on the analysis and verification of the collected information from the sales team and the warehouse team.P24-016PCT1

[0071] As may be understood, the process undertaken by the first user 104(1), the sales team, and the warehouse team to work on improving the performance of the organization 108 is tedious, time-consuming, and resource-intensive. Even if conventional Al agents are used in place of the sales team and the warehouse team to automate their responsibilities, the process may face compatibility issues. This process can also be time-consuming and resourceintensive. Also, when such Al agents are used, the chances of the results obtained from each Al agent being imprecise are higher. Imprecise results can often lead to providing inaccurate solutions to various problems faced by the first user 104(1) in the organization 108, thereby negatively impacting the performance of the organization 108. Further, since such requirements are quite common, the first user 104(1) and other teams in the organization 108 may have to perform the same process again and again, leading to inefficient use of the time of the first user 104(1) and other teams in the organization 108. Furthermore, the first user 104(1) may have to log on to a particular portal if one or more software tools are used to get a solution to a particular problem. Since many users may simultaneously wish to utilize these Al agents and may log on to the portal and spend time on it throughout the day, the portal itself gets burdened leading to higher processing and memory requirements, which in turn leads to higher infrastructure and electricity costs.

[0072] To overcome these problems, an approach for generating a response which is an accurate solution to a problem stated through a query is provided. To that end, the present disclosure describes the server system 102 which performs a plurality of operations for facilitating the generation of the expected response to the query.

[0073] In one embodiment, the environment 100 may further include a database 112 coupled with the server system 102. In an example, the server system 102 coupled with the database 112 is embodied within a central server (not shown) associated with the employer of the first users 104, however, in other examples, the server system 102 can be a standalone component (acting as a hub) connected to the central server.

[0074] The database 112 may be incorporated in the server system 102 or maybe an individual component connected to the server system 102 or maybe a database stored in cloud storage. In one embodiment, the database 112 may store one or more Al agents, such as a master Al agent 114 and a plurality of available domain-specific Al agents 116(1), 116(2), . . . 116(N) (hereafter collectively referred to as ‘available domain-specific Al agents 116’ or ‘domain-specific Al agents 116’). Herein, ‘N’ is a non-zero natural number. In anotherP24-016PCT1embodiment, the database 112 may also store other necessary machine instructions required for implementing the various functionalities of the server system 102 such as firmware data, operating system, and the like. It is noted that the master Al agent 114 and the domain-specific Al agents 116 have been explained in detail later in the present disclosure. In addition, the database 112 provides a storage location for data and / or metadata obtained from various operations performed by the server system 102.

[0075] In an embodiment, when a query is prompted by a stakeholder such as the first user 104(1), the server system 102 is configured to receive the query. More specifically, the server system 102 is configured to transfer the query to the master Al agent 114. The master Al agent 114 may have to understand a predefined context associated with a query. Upon understanding the predefined context, the intent associated with the query may be determined based on the predefined context. Further, a set of operations may be performed based on the intent to generate the response which is explained later in the present disclosure.

[0076] As may be understood, the predefined context of the query refers to a specific set of circumstances and background information that influences or surrounds the query. The predefined context includes factors that help the master Al agent 114 to understand the intent, scope, and relevance of the query. More specifically, information type in the predefined context includes user persona, location, time, prior interactions, user preferences, ongoing activities, external data (e.g., weather, calendar events, etc.), or the like. For instance, if the first user 104(1) is the manager of the organization 108, then the master Al agent 114 may understand that the individual asking the query is well aware of the terminologies and criteria that are internal to the organization 108. Further, accordingly, the master Al agent 114 may generate a response that may be well understandable to the manager. On the other hand, if the first user 104(1) is a customer receiving services from the organization 108, then the master Al agent 114 may have to, initially, verify the identity of the customer, authenticate the customer, check for constraints that have to be applied to the customer, and the like, before responding to the query. Herein, the constraints can include checking whether the individual is allowed to ask such queries, whether the individual is even authorized to ask any queries, whether the individual is a fraudster, etc. Further, once it is identified that the individual is an authorized individual, the master Al agent 114 may have to search for a portion of the database 112 that is assigned to this particular customer and then check for the different organizational processes that may be related to this particular customer. Finally, based on these details, the master Al agent 114 may generate a relevant response for the query.P24-016PCT1

[0077] In one embodiment, the predefined context of the query is defined while designing the master Al agent 114 and hence the master Al agent 114 is pre-fed with the context. For instance, a chatbot linked to the master Al agent 114 is designed to solve problems associated with a particular organizational process of the organization 108 such as a sales problem, then the chatbot will provide solutions for sales-related problems only.

[0078] In another embodiment, the master Al agent 114 may be designed to be more generic, i.e., it is designed to understand different contexts that may be associated with the query. Once the context is understood, then based on the context, the intent of the query may be determined to generate an accurate response to the query. For instance, a chatbot linked to the master Al agent 114 is designed to provide solutions to different problems associated with different organizational processes of the organization 108. In such a scenario, upon receiving a query, the master Al agent 114 may have to understand the context, i.e., whether it is related to a sales process, a warehouse process, a distribution process, or the like. The step of determining the context can also include understanding the persona of the individual asking the query. In some embodiments, the query can be associated with multiple contexts. For instance, a query requesting information related to both the sales process and the warehouse process.

[0079] In one embodiment, the server system 102 is configured to process textual data associated with the query using the master Al agent 114 for understanding the predefined context of the query. As may be understood, the master Al agent 114 includes sensors, processing units, memory, and actuators that operate together in an environment. Thus, it can be stated that the master Al agent 114 is an engine that can collect data from the environment through sensors or direct human prompts, process the data by analyzing it, recognize patterns, and make sense of the data. Based on the processed information, the master Al agent 114 makes decisions, selects actions to be taken, and then performs the selected actions on the agent’s environment. In a non-limiting implementation, the master Al agent 114 is associated with one or more ML models used for learning and decision-making. Examples of the ML models include Convolutional Neural Network (CNN)-based models, Recurrent NN (RNN)- based models, Long Short-Term Memory (LSTM)-based models, Reinforcement Learning (RL)-based models, transformer-based models, Large Language Models (LLMs), Generative Al models, ensemble models, and the like. It may be understood that the ML models are used during learning and decision-making steps. During this process, the context of the query may be determined if not pre-defined and then the intent of the query is determined. Thus, it mayP24-016PCT1be understood, that in one implementation, as analysis of text is involved, Natural Language Processing (NLP) can be used to determine the context of the query. The process of determining the context and / or understanding the context associated with the query includes applying NLP, intention recognition, contextual analysis and understanding, context memory update, feedback, and learning. The process of determining the context using NLP is a well- known process to a person skilled in the art and hence is not elaborated on in the present disclosure.

[0080] Upon understanding the predefined context of the query, the master Al agent 114 understands the intent associated with the query. In one embodiment, the master Al agent 114 is configured to determine the intent of the query based, at least in part, on the predefined context of the query and historical intent information that is accessible to the master Al agent 114. As may be understood, the intent refers to the underlying reasoning of the query, i.e., understanding what exactly the first user 104(1) is trying to ask. In one embodiment, the intent can include a single intent. In another embodiment, the intent can include multiple intents, such as a first intent, a second intent, or the like. For instance, the query is: " provide a performance of a warehouse F . Herein, the master Al agent 114 may have to understand that this query is associated with just a single intent. The master Al agent 114 may also have to understand that the first user 104(1) is trying to understand how well warehouse Y is performing and whether the performance has reduced or increased. Alternatively, assuming the query is: "provide the performance of the warehouse Y and details of a customer whose items are stored in if . Herein, the master Al agent 114 determines that there are two intents in the query. One is to understand the performance of warehouse Y and the second intent is to fetch details of customers whose items the warehouse Y is holding. Herein, the overall intent of the first user 104(1) can be to understand a portfolio of a particular customer whose items are stored in the warehouse Y and check if the performance of the warehouse Y and the items of the corresponding customer have an impact on each other.

[0081] More specifically, to determine the intent, the server system 102 is configured to process the textual data associated with the query using the master Al agent 114 based on the predefined context. In one embodiment, the master Al agent 114 is configured to apply NLP and / or Natural Language Understanding (NLU), to categorize the query into one or more intents, and identify entities within the query. Herein, the ML models such as LLM or Generative Al can be used to categorize the intent more efficiently. More specifically, processing of the textual data associated with the query to determine the intent include textP24-016PCT1preprocessing, feature extraction, model training, intent classification using the trained model, using advanced techniques (e.g., contextual understanding, sequence determination models, attention mechanisms, etc.), and the like. Herein, it is noted that the model may be trained to understand the intent associated with the query based on the predefined context and the historical intent information. In one embodiment, the historical intent information includes past user queries that may have been received from the first user 104(1) and the second user 106(1). The historical intent information may also include labeled intent data including historical datasets where the user queries are labeled with their corresponding intents, user interaction history, contextual information surrounding the past queries, conversational history, feedback and corrections, multimodal data, domain-specific data, and the like. The process of determining the intent based on the predefined context and the historical intent information using the ML models is also well known to a person skilled in the art, and hence it is not repeated herein for the sake of brevity.

[0082] It is noted that determining the context and the intent associated with the query for generating an appropriate response are closely interrelated steps. Ideally, knowing the context while determining the intent is preferred as it helps disambiguate the query. However, if the query is not ambiguous, then for determining the intent and generating the response to the query, context may not be required. Further, in some scenarios, the intent can be determined without being aware of the context as well. For instance, if the query is: ‘provide performance of a customer", herein the intent can be determined which is to determine performance of a customer who is receiving services from the organization 108. However, in this query, the context is not clear and is also not available with the master Al agent 114, i.e., the customer identity, period of performance, and performance parameters are not clear. Therefore, the master Al agent 114 may initially determine the intent by processing the textual data associated with the query using the above-mentioned techniques. Further, in one embodiment, the master Al agent 114 generates a broader response. For instance, for the query ‘provide a performance of a customer" , the master Al agent 114 provides information about all performance parameters related to all the customers of the organization 108. In an alternative embodiment, the master Al agent 114 may ask follow-up questions to the first user 104(1) to understand the context. For instance, for the query, ‘provide a performance of a customer" , the follow-up questions can include asking the user to specify the identity of the customer and specify the performance parameters that the user has asked for. Once the context is received by the master Al agent 114, the intent may be precise, resulting in generating a precise response.P24-016PCT1

[0083] In some embodiments, the intent and the context may be determined parallelly, and a dependency can be established between the two, refining the understanding of both the intent and the context, iteratively. Also, the context may be updated as more information becomes available to the master Al agent 114, thereby refining the intent and the response for the query.

[0084] As may be understood, Al agents are designed for a specific purpose, such as an Al agent to provide customer service, an Al agent for content creation, an Al agent for online shopping, etc. Moreover, the basic operation of an Al agent is to search, compare, and return results. Thus, a single Al agent can perform these operations for a specific goal or task associated with a single purpose. To that note, the master Al agent 114 is configured to operate a set of domain-specific Al agents 116(1), 116(2), . . . 116(J) (hereafter, collectively referred to as ‘set of domain-specific Al agent 116(1)-116(J)’ or ‘domain-specific Al agent 116(1)- 116(J)’) to generate an insight, based at least on the intent. In other words, the master Al agent 114, upon understanding the intent of the query, searches for the domain-specific Al agents 116(1)-116(J) that are more suitable for the intent, facilitates the domain-specific Al agents 116(1)-116(J) to search, compare and return an insight as a result to the master Al agent 114. Then, the master Al agent 114 compares the insights obtained from each of the domain-specific Al agents 116(1)-116(J) with corresponding expected results and generates a final and consolidated response to the query. Herein, ‘J’ is a nonzero natural number that is less than or equal to ‘N’ . Moreover, it may be understood that the set of domain-specific Al agents 116(1)- 116(J) is a subset of the domain-specific Al agents 116.

[0085] Thus, in one embodiment, the server system 102 is configured to select the set of domain-specific Al agents 116(1 )- 116(J) for the intent based, at least in part, on one or more relevancy factors between the set of domain-specific Al agents 116(1)-116(J) and the intent. In a specific embodiment, the master Al agent 114 is communicatively coupled to the set of domain-specific Al agents 116(1)-116(J). The process of selecting the set of domain-specific Al agents 116(1)-116(J) is explained later in the present disclosure.

[0086] Further, the server system 102 may be configured to determine a activity execution operation to be performed based, at least in part, on the intent associated with the query and query execution information for the intent. In one embodiment, the server system 102 determines the activity execution operation using the set of domain-specific Al agents 116(1)-116(J). More specifically, each of the set of domain-specific Al agents 116(1)-116(J)P24-016PCT1selected by the master Al agent 114 determines the activity execution operation to be performed to generate an insight into the intent of the query. The process of determining the activity execution operation is also explained later in the present disclosure.

[0087] Upon determining the activity execution operation to be performed, each of the set of domain-specific Al agents 116(1)-116(J) may implement or perform the activity execution operation. The process of implementing the activity execution operation is also explained later in the present disclosure. Upon implementation of the activity execution operation, a set of resolutions may be obtained from each of the set of domain-specific Al agents 116(1)-116(J). The server system 102 may be configured to identify the set of resolutions for the activity execution operation through the set of domain-specific Al agents 116(1)- 116(J). Herein, each of the set of resolutions is indicative of an outcome of performing a particular activity associated with the activity execution operation.

[0088] Further, the server system 102 may be configured to determine an insight based on the set of resolutions from each of the domain-specific Al agents 116(1)-116(J). Thus, it may be understood that each domain-specific Al agent ( / .< ., each of the domain-specific Al agents 116(1)-116(J)) determines an insight from the set of resolutions generated by the corresponding domain-specific Al agent. In one embodiment, the insight from each of the domain-specific Al agents 116(1)-116(J) is provided to the master Al agent 114. Thus, the server system 102, through the master Al agent 114 may be configured to generate a response to the query based on the insight determined by each of the domain-specific Al agents 116(1)- 116(J).

[0089] Although in FIG. 1, the server system 102 is shown to be incorporated within the environment 100, in some embodiments, the server system 102 may be external to and in communication with the environment 100, for example, via the network 110. In some examples, the server system 102 may be implemented in third-party external servers to perform the various operations described herein.

[0090] The number and arrangement of systems, devices, and / or networks shown in FIG. 1 are provided as an example. There may be additional systems, devices, and / or networks; fewer systems, devices, and / or networks; different systems, devices, and / or networks; and / or differently arranged systems, devices, and / or networks than those shown in FIG. 1. Furthermore, two or more systems or devices shown in FIG. 1 may be implemented within a single system or device, or a single system or device is shown in FIG. 1 may be implementedP24-016PCT1as multiple, distributed systems or devices. In addition, the server system 102 should be understood to be embodied in at least one computing device in communication with the network 110, which may be specifically configured, via executable instructions, to perform steps as described herein, and / or embodied in at least one non-transitory computer-readable media.

[0091] FIG. 2 illustrates a simplified block diagram of a server system 200, in accordance with an embodiment of the present disclosure. It is noted that the server system 200 may be similar to the server system 102 of FIG. 1. In one embodiment, the server system 200 is a part of the internal server operated by an organization (e.g., the organization 108) employing stakeholders for managing different organizational processes such as the first users 104. In some embodiments, the server system 200 is embodied as a cloud-based and / or Software as a Service (SaaS) based architecture.

[0092] The server system 200 includes a computer system 202 and a database 204. It is noted that the database 204 is identical to the database 112 of FIG. 1. The computer system 202 includes at least one processor 206 (herein, referred to interchangeably as ‘processor 206’) for executing instructions, a memory 208, a communication interface 210, a user interface 212, and a storage interface 214 that communicates with each other via a bus 216.

[0093] In some embodiments, the database 204 is integrated into the computer system 202. For example, the computer system 202 may include one or more hard disk drives like the database 204. The user interface 212 is an interface, such as a Human Machine Interface (HMI) or a software application that allows users such as an administrator to interact with and control the server system 200 or one or more parameters associated with the server system 200. It may be noted that the user interface 212 may be composed of several components that vary based on the complexity and purpose of the application. Examples of components of the user interface 212 may include visual elements, controls, navigation, accessibility features, etc.

[0094] The storage interface 214 is any component capable of providing the processor 206 with access to the database 204. The 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 Redundant Array of Independent Disc (RAID) controller, a Storage Area Network (SAN) adapter, a network adapter, and / or any component providing the processor 206 with access to the database 204.

[0095] In one non-limiting example, the database 204 is configured to storeP24-016PCT1organization information 218, agent profile information 220, a master Al agent 222, a plurality of available domain-specific Al agents 224(1), 224(2), ... , 224(N) (hereafter collectively referred to as ‘available domain-specific Al agents 224’ or ‘domain-specific Al agents 224’), a Large Language Model (LLM) 226, a query-response dataset 228, and the like. Herein, the master Al agent 222 and the domain-specific Al agents 224 are similar to the master Al agent 114 and the domain-specific Al agents 116 of FIG. 1.

[0096] In one embodiment, the organization information 218 includes information related to various factors associated with the organization 108. In a non-limiting example, the factors include organization levels, organizational processes, technology, stakeholders, customers, and the like. More specifically, the organization information 218 may include organization-level management information including the overall goal, mission, or vision of the organization, distribution of responsibilities among different stakeholders in the organization at different organizational levels (e.g., top-level, mid-level, and frontline level), marketing and sales, a track of profitability over the years, and the like. Further, the organization information 218 also includes organizational process-related information including different organizational processes, such as product design process, quality control, sales process, marketing process, finance management process, compliance and regulatory process, and the like. The organizational process-related information may also include different activities that may be associated with each organizational process, predefined activity execution information, stakeholders allotted to each organizational process, one or more tools that are required for the completion of each organizational process, and the like. In a specific embodiment, the predefined activity execution information includes query execution information for different intents that may be associated with the query, query execution information for different contexts that may be associated with the query, predefined category information, task-related information, predefined goal execution information, and the like. Herein, the predefined activity execution information includes information indicating different steps that need to be implemented for the execution of a particular activity, such as the query, task, goal, or the like. The organization information 218 may also include historical information such as historical intent information, historical context information, and the like.

[0097] Further, in one embodiment, the agent profile information 220 includes information related to profiles corresponding to the domain-specific Al agents 224. The profile corresponding to a particular domain-specific Al agent (e.g., the domain-specific Al agent 224(1)) is generated based on the information, such as, but not limited to, types of tasks thatP24-016PCT1the domain-specific Al agent 224(1) is best suited for, performance metrics, historical performance data including previous task outcomes and success rate, current state of the agent, and the like.

[0098] Furthermore, in one embodiment, the query-response dataset 228 includes historic queries and their corresponding responses. It is noted that, upon generating the responses to queries that may be received in real-time or in the future, the query-response dataset 228 may be updated with the latest queries and their corresponding responses.

[0099] The processor 206 includes suitable logic, circuitry, and / or interfaces to execute operations for determining a activity execution operation for addressing the query, identifying a set of resolutions for the activity execution operation, determining an insight for the set of resolutions, generating a response to the query, and the like. Examples of the processor 206 include, but are not limited to, an Application-Specific Integrated Circuit (ASIC) processor, a Reduced Instruction Set Computing (RISC) processor, a Graphical Processing Unit (GPU), a Complex Instruction Set Computing (CISC) processor, a Field-Programmable Gate Array (FPGA), and the like.

[0100] The memory 208 includes suitable logic, circuitry, and / or interfaces to store a set of computer-readable instructions for performing the various operations described herein. Examples of the memory 208 include a random-access memory (RAM), a read-only memory (ROM), a removable storage drive, a hard disk drive (HDD), and the like. It will be apparent to a person skilled in the art that the scope of the disclosure is not limited to realizing the memory 208 in the server system 200, as described herein. In another embodiment, the memory 208 may be realized in the form of a database server or a cloud storage working in conjunction with the server system 200, without departing from the scope of the present disclosure.

[0101] The processor 206 is operatively coupled to the communication interface 210, such that the processor 206 is capable of communicating with a remote device ( / .< ., to / from a remote device 230) such as third-party servers or with electronic devices associated with the first users 104 or communicating with any entity connected to the network 110 (as shown in FIG. 1).

[0102] It is noted that the server system 200 as illustrated and hereinafter described is merely illustrative of an apparatus that could benefit from embodiments of the present disclosure and, therefore, should not be taken to limit the scope of the present disclosure. It is noted that the server system 200 may include fewer or more components than those depictedP24-016PCT1in FIG. 2.

[0103] In one implementation, the processor 206 includes a master-slave operation management module 232, an insight generation module 234, an evaluator module 236, and an insight validation module 238. It should be noted that components, described herein, such as the master-slave operation management module 232, the insight generation module 234, the evaluator module 236, and the insight validation module 238 can be configured in a variety of ways, including electronic circuitries, digital arithmetic, and logic blocks, and memory systems in combination with software, firmware, and embedded technologies. In various non-limiting examples, the master-slave operation management module 232, the insight generation module 234, the evaluator module 236, and the insight validation module 238 may be communicatively coupled with each other and can transfer data between the modules.

[0104] In an embodiment, the master-slave operation management module 232 includes suitable logic and / or interfaces for receiving the query from the first user (e.g., the first user 104(1)). Upon receiving the query, the master-slave operation management module 232 may access the master Al agent 222 from the database 204. Further, the master-slave operation management module 232 may facilitate the master Al agent 222 to determine the intent associated with the query. Thus, in one embodiment, the master Al agent 222 determines the intent associated with the query based, at least in part, on the predefined context of the query and the historical intent information of the master Al agent 222. The process of determining the intent has been already explained with reference to FIG. 1. To that end, the same is not explained again for the sake of brevity. Upon determining the intent of the query, the master-slave operation management module 232 may be configured to receive the determined intent from the master Al agent 222 and store it in the database 204 for the corresponding query which is accessible for future use by any module of the server system 200.

[0105] In another embodiment, the master-slave operation management module 232 is configured to facilitate the master Al agent 222 to select a set of domain-specific Al agents 224(1 )-224(J) for the intent based, at least in part, on the relevancy factors between the set of domain-specific Al agents 224(1)-224(J) and the intent. Herein, the master Al agent 222 is configured to operate the set of domain-specific Al agents 224(1 )-224(J) to generate the insight, based at least on the intent. Herein, ‘J’ is a non-zero natural number that is less than or equal to ‘N’. Moreover, the domain-specific Al agents 224(1)-224(J) are a subset of theP24-016PCT1domain-specific Al agents 224 and are similar to the domain-specific Al agents 116(1 )- 116(J) of FIG. 1. Also, it may be understood that the relation between the master Al agent 222 and the domain-specific Al agents 224 is a master-slave relationship, and hence every operation of the domain-specific Al agents 224 is controlled based on receiving control signals from the master Al agent 222.

[0106] In addition, the master-slave operation management module 232 is configured to control the master-slave relationship by verifying whether master-slave guidelines are followed during the operation of the domain-specific Al agents 224(1)-224(J) by the master Al agent 222. In a non-limiting example, the master-slave guidelines may include establishing well-defined communication protocols between the master and slave agents, establishing proper synchronization in time-sensitive operations, implementation of fault tolerance and robust error handling, scalability, security, efficient task assignment and coordination, monitoring and feedback, modularity and maintainability, consistency and data integrity, compliance and standards, and the like.

[0107] Further, to select a particular domain-specific Al agent such as the domainspecific Al agent 224(1) in the set of domain-specific Al agents 224(1)-224(J) for the intent, the master Al agent 222 may be configured to access a plurality of agent profiles of the plurality of available domain-specific Al agents 224. The master Al agent 222 may further be configured to calculate a plurality of relevancy factors between the plurality of agent profiles and the intent. Further, the master Al agent 222 may select the domain-specific Al agent (e.g., the domain-specific Al agent 224(2)) associated with an agent profile that has the highest relevancy factor from the plurality of relevancy factors. The term ‘relevancy factor’ refers to a factor or a metric that indicates a suitability extent between two entities. For instance, if an agent profile of a domain-specific Al agent 224(2) is highly suitable to perform tasks for generating relevant details to respond to a query with a particular intent, the corresponding domain-specific Al agent 224(2) and the intent have a large value for the relevancy factor between them.

[0108] In a specific embodiment, the relevancy factor may be associated with matching criteria. The matching criteria include a set of criteria that facilitate the calculation of the relevancy factor. In a non-limiting example, the matching criteria may include performance requirements such as matching required accuracy and speed with the capabilities of the domain-specific Al agents 224. The matching criteria may further include resource constraints,P24-016PCT1availability, specialization, and the like. A threshold may be set for each criterion in the matching criteria. For example, the specialization, i.e., the purpose for which a particular domain-specific Al agent is designed, should match the intent of the query. That particular domain-specific Al agent should not be loaded with another task. It should have sufficient resources such as memory and processing power required for the processing tasks to meet the intent, and the performance requirements should also be matched to the intent. Once a particular domain-specific Al agent meets all these matching criteria with a particular intent of the query, then a relevancy factor of the corresponding domain-specific Al agent is computed to be relevant for the intent. Further, that particular domain-specific Al agent may have to be selected for the corresponding intent.

[0109] To that note, the master Al agent 222 may examine the matching criteria and compute the relevancy factor for selecting a domain-specific Al agent for the intent. In one embodiment, the master Al agent 222 performs these operations using a rule-based selection approach. In another embodiment, the master Al agent 222 employs a weighted scoring operation in which scores are assigned to each domain-specific Al agent and one with the highest score can be selected for performing tasks for the corresponding intent. In yet another embodiment, the master Al agent 222 employs one or more ML models that are trained to predict a best-fit domain-specific Al model for the determined intent based on historical data and the intent.

[0110] Upon selecting the set of domain-specific Al agents 224(1 )-224(J), the masterslave operation management module 232 may be configured to access the corresponding set of domain-specific Al agents 224(1)-224(J) from the database 204. Further, the master-slave operation management module 232 may be configured to transfer the query or a portion of the query associated with the determined intent to the insight generation module 234.

[0111] In one embodiment, the insight generation module 234 includes suitable logic and / or interfaces for receiving the query or the portion of the query with determined intent. The insight generation module 234 may be configured to transfer this query or the corresponding portion of the query with the intent to the corresponding set of domain-specific Al agents 224(1 )-224(J) for further processing. Herein, if the determined intent corresponds to a single intent, then the query is transferred to the set of domain-specific Al agents 224(1)- 224(J). In one embodiment, the set of domain-specific Al agents 224(1)-224(J) can include a single domain-specific Al agent (e.g., the domain-specific Al agents 224(1)). In anotherP24-016PCT1embodiment, the set of domain-specific Al agents 224(1 )-224(J) can include more than one domain-specific Al agents 224(1)-224(J), with each of the set of domain-specific Al agents 224(1 )-224(J) partially contributing to determining a set of resolutions for the intent.

[0112] In a specific embodiment, each of the set of domain-specific Al agents 224(1)- 224(J) is configured to determine an activity execution operation to be performed based, at least in part, on the intent associated with the query and query execution information for the intent. In another embodiment, each of the set of domain-specific Al agents 224(1)-224(J) is configured to identify a set of resolutions for the activity execution operation. Herein, each of the set of resolutions is indicative of an outcome of performing a particular activity associated with the activity execution operation by each of the set of domain-specific Al agents 224(1)- 224(J). In yet another embodiment, each of the set of domain-specific Al agents 224(1)-224(J) is configured to generate an insight based, at least in part, on the set of resolutions. Thus, it may be understood that the insight generation module 234 is configured to generate the insight for the intent through the set of domain-specific Al agents 224(1 )-224(J) that are selected by the master Al agent 222 for the intent associated with the query.

[0113] In a non-limiting implementation, for generating the activity execution operation, the insight generation module 234 is configured to identify a query category for the intent based, at least in part, on predefined category information. In one embodiment, the query category includes one of a task, a goal, or a combination thereof. Thus, it may be understood that, for a particular intent, the query can either include a task, a goal, or both. Based on the query category, the activity execution operation may be determined.

[0114] In one embodiment, upon identifying the query category to be the task, the insight generation module 234 is configured to generate dynamically, a task execution operation to be performed for generating the insight based, at least in part, on the intent and task-related information. If the task execution operation requires splitting the task into subtasks, then the task execution operation can include a set of sub-tasks to be performed for generating the insight. The insight generation module 234 may assign the task execution operation to the activity execution operation to be performed. In various examples, the task- related information includes different types of tasks, examples of tasks and example processes to complete the tasks, examples of sub-tasks that may be associated with example tasks, task identifiers (IDs), resolutions obtained from the execution of the example tasks, and the like. In various other examples, the task-related information includes task execution patterns, taskP24-016PCT1execution tools, access credentials for such tools, information about all possible tasks that may be required to be performed for existing organizational processes in the organization 108, and the like.

[0115] Thus, it may be understood that the insight generation module 234 may determine the task execution operation for the task by comparing the intent with the task- related information. More specifically, the insight generation module 234 may facilitate the set of domain-specific Al agents 224(1 )-224(J) to generate dynamically, the task execution operation. For instance, the query is: ‘ What is the contact number of a customer X? The master Al agent 222 understands the intent and selects the most suitable domain-specific Al agent such as the domain-specific Al agent 224(1). The master Al agent 222 transfers the query to the domain-specific Al agent 224(1). The domain-specific Al agent 224(1), then identifies the query category of the query to be a task. The task is to search the contact number of customer X from a portion of the database 204 assigned to the corresponding customer X. The insight generation module 234 checks the task-related information to determine the steps that need to be performed to generate a resolution for the task. The insight generation module 234 generates the task execution operation that can include a set of sub-tasks if needed. For this particular example, the domain-specific Al agent 224(1) is made to perform the set of subtasks, such as searching for the name of the customer X and then retrieving all the contact numbers associated with the customer X. Herein, the set of resolutions includes a resolution including all the contact numbers of the customer X. Further, the insight may be generated as follows:

[0116] In another embodiment, upon identifying the query category to be the goal, the insight generation module 234 is configured to determine a goal execution operation to be performed for determining the insight based, at least in part, on the intent and predefined goal execution information. The insight generation module 234 may then assign the goal execution operation to the activity execution operation to be performed. As may be understood, the goals in any organization are associated with a predefined set of steps that need to be performed toP24-016PCT1a solution to the corresponding goals. Thus, in various examples, the predefined goal execution information includes example goals, tasks associated with the example goals that need to be performed to provide resolutions to the example goals, historic goals, associated tasks, and corresponding resolutions, goal execution criteria, tools required for the execution of the goals, tool access credentials, example goal execution operations, the goal execution operation, and the like. More specifically, the insight generation module 234 may facilitate the set of domainspecific Al agents 224(1)-224(J) to determine the goal execution operation. The goal execution operation can include a set of tasks to be performed that are predefined.

[0117] For instance, if the query is: provide a summary of a customer , then the master Al agent 222 understands the intent and selects the most suitable domain-specific Al agent such as the domain-specific Al agent 224(2). The master Al agent 222 transfers the query to the domain-specific Al agent 224(2). The domain-specific Al agent 224(2), then identifies the query category of the query to be a goal. Since the goal has predefined steps, the insight generation module 234 may search for the goal in the predefined goal execution information from the database 204. Upon searching, the following set of tasks may appear which are to be performed to obtain resolutions for the goal:

[0118] The domain-specific Al agent 224(2) may implement this set of tasks using one or more tools including SQL tools, semantic tools, data frames, native tools, one or more ML models, or the like. Upon implementation of each task, a resolution may be obtained. In one embodiment, the resolution may be provided as input to the next task or one or more remaining tasks of the set of tasks. Then, upon implementation of all the tasks in the set of tasks, the set of resolutions including either a single resolution or multiple resolutions may be generated. In an alternative embodiment, the resolution of each task in the set of tasks of the goal may be taken individually as the set of resolutions. The insight generation module 234 may generate an insight from the set of resolutions using the domain-specific Al agent 224(2). The insight may appear as follows:P24-016PCT1

[0119] Further, in yet another embodiment, upon identifying the query category to be a combination of the task and the goal, the insight generation module 234 is configured to segregate the query into at least two sub-queries. The two sub-queries may include a first subquery indicative of the task and a second sub-query indicative of the goal. The insight generation module 234 may further be configured to generate and assign dynamically, a task execution operation to the activity execution operation for generating the insight for the first sub-query. Further, the insight generation module 234 may determine a goal execution operation to the activity execution operation for generating the insight for the second subquery. More specifically, the insight generation module 234 may facilitate the set of domainspecific Al agents 224(1 )-224(J) to segregate the query and generate the set of sub-tasks for the first sub-query and the set of tasks for the second sub-query.

[0120] For instance, the query is: ‘ Provide a summary of a customer X and what is the average unsold space a region Y in Feb 2023' . Then, the master Al agent 222 understands the intent and selects the most suitable domain-specific Al agents such as the domain-specific Al agents 224(1) and 224(3). The master Al agent 222 transfers the right portion of query to domain-specific Al agents 224(1) and 224(3). First portion ‘Provide a summary of a customer X’ to agent 224(1) and second position ‘what is the average unsold space a region Y in Feb 2023’ to agent 224(3). Each of the domain-specific Al agents 224(1) and 224(3), then identifies the query category of the query. Herein, the domain-specific Al agent 224(1) is designed for the management of the sales process in the organization 108. Thus, the domain-specific Al agent 224(1) understands that the query asking for the summary of the customer X is a goal and hence, the domain-specific Al agent 224(1) may generate the set of tasks as follows:P24-016PCT1

[0121] The domain-specific Al agent 224(1) may generate the set of resolutions upon the implementation of the set of tasks. Finally, based on the set of resolutions, the domainspecific Al agent 224(1) may generate the insight which is as follows:

[0122] Parallelly, the domain-specific Al agent 224(3), as it is designed for the management of the marketing process of the organization 108, understands that the query is a task which is to determine the average unsolved space in the region Y for the corresponding month and the year. In this example, the domain-specific Al agent 224(3) need not have to generate a set of sub-tasks for the implementation of the tasks. The domain-specific Al agent 224(3) may directly implement the task and obtain the resolution. Based on the resolution, an insight may be generated based on the resolution. Thus, the insight generated is as follows:P24-016PCT1

[0123] The insight generation module 234 may generate the insight from the set of resolutions at each of the set of domain-specific Al agents 224(1)-224(J) and provide it to the master Al agent 222. In one embodiment, the insight can be a summary that the domainspecific Al agent 224(1) to whom the intent is assigned may be generated upon processing and understanding the set of resolutions. In another embodiment, the insight can be a consolidation of the set of resolutions. In yet another embodiment, the insight can be an understanding obtained by the domain-specific Al agent 224(1) from the set of resolutions. However, there is a possibility that the insight generated from the set of resolutions is inaccurate due to the inaccuracies in the activity execution operation generated by the corresponding set of domainspecific Al agents 224(1)-224(J). Thus, there is a requirement to evaluate the accuracy or validity of the activity execution operation, before transferring the insights from each of the set of domain-specific Al agents 224(1)-224(J) to the master Al agent 222. To that note, the activity execution operation, the set of resolutions, the insight, and the intent are transferred to the evaluator module 236.

[0124] In one embodiment, the evaluator module 236 includes suitable logic and / or interfaces for determining a validity of the activity execution operation based, at least in part, on the activity execution operation, the set of resolutions, and the LLM 226. In various examples, the LLM can be a fine-tuned LLM including a Bidirectional Encoder Representations from Transformers (BERT) model, a Generative Pre-trained Transformer (GPT) model, or the like. For determining the validity, the evaluator module 236 may be configured to access the activity execution operation and the set of resolutions. In a specific embodiment, the evaluator module 236 is configured to control the operation of an evaluator such as the LLM 226 associated with each of the set of domain-specific Al agents 224(1)- 224(J). Thus, it may be understood that the LLM 226 associated with each of the set of domainspecific Al agents 224(1)-224(J) is configured to determine the validity of the activity execution operation generated by the corresponding domain-specific Al agent of the set of domain-specific Al agents 224(1)-224(J). For example, the LLM 226 associated with a first Al agent such as the domain-specific Al agent 224(1) evaluates the validity of the set of tasks generated by the corresponding first Al agent. Similarly, the LLM 226 associated with a second Al agent such as the domain-specific Al agent 224(2) evaluates the validity of the set of tasks generated by the corresponding second Al agent.P24-016PCT1

[0125] Further, the evaluator module 236 may transfer the accessed set of tasks and the set of resolutions to the evaluator (e.g., the LLM 226) for the LLM 226 to determine the validity of the activity execution operation. Furthermore, the evaluator module 236 may determine a set of fine-tuning parameters for each activity of the activity execution operation based, at least in part, on the intent and a predefined training dataset of the LLM 226. In a non-limiting implementation, the evaluator module 236 may determine the set of fine-tuning parameters using the LLM 226.

[0126] As may be understood, an LLM is generally trained using a diverse and extensive dataset for language understanding. This dataset covers a wide range of topics, styles, and languages to ensure the model’s generalization ability. However, when the LLM 226 is used for determining the validity of the activity execution operation generated by a particular domain-specific Al agent associated with the particular LLM 226, then the LLM 226 is finetuned based on the predefined training dataset. In an embodiment, the evaluator module 236 is configured to train the LLM 226 using Reinforcement Learning with an Al Feedback (RLAIF)- based technique. This technique aligns LLMs with human preferences and ethical considerations.

[0127] In a non-limiting example, the predefined training dataset includes information specific to the domain of the corresponding domain-specific Al agent. Herein, the domain corresponds to one of the various organizational processes of the organization 108. More specifically, the predefined training dataset may include information, such as the details related to various data sources associated with a particular domain-specific Al agent, intent for which the domain-specific Al agent is designed, tasks that can be generated for corresponding intent, historical operations associated with a corresponding domain-specific Al agent, and the like.

[0128] Moreover, the evaluator module 236 transfers the set of fine-tuning parameters to the domain-specific Al agent based at least on the intent. In a non-limiting implementation, the evaluator module 236 may transfer the set of fine-tuning parameters to the domain-specific Al agent through the LLM 226 associated with the corresponding domain-specific Al agent. The domain-specific Al agent associated with the LLM 226 may be configured to receive the set of fine-tuning parameters. Upon receiving the set of fine-tuning parameters, in one embodiment, the domain-specific Al agent may be configured to identify the activity execution operation to be valid when the set of fine-tuning parameters matches with fine-tuning criteria. In another embodiment, the domain-specific Al agent may be configured to identify theP24-016PCT1activity execution operation to be invalid when the set of fine-tuning parameters deviates from the fine-tuning criteria.

[0129] In a non-limiting implementation, upon identifying that the activity execution operation is invalid, the evaluator module 236 is configured to facilitate the domain-specific Al agent to generate a new activity execution operation based, at least in part, on the intent and the set of fine-tuning parameters. The domain-specific Al agent may further be configured to identify a new set of resolutions for the new activity execution operation based, at least in part, on performing the new activity execution operation. Further, the evaluator module 326 may be configured to determine a validity of the new activity execution operation based, at least in part, on the new activity execution operation, the new set of resolutions, and the LLM 226. It is noted that, if the new activity execution operation is also determined to be invalid, then another new activity execution operation may be generated, and their validity may be checked again. Thus, it may be understood that the process of generating the new activity execution operation and determining the validity of these activities may be performed iteratively until the validity of the activities to be performed is determined to be valid by the evaluator module 326. The process of determining the validity of the new activity execution operation is similar to the process explained earlier for the activity execution operation. Moreover, the process of determining the set of fine-tuning parameters for fine-tuning the activity execution operation and the new activity execution operation is explained later in the present disclosure.

[0130] Upon determining the activity execution operation or the new activity execution operation that are valid, the set of resolutions and the insight from the corresponding activity execution operation are determined for each of the set of domain-specific Al agents 224(1)- 224(J). The insight from each of the set of domain-specific Al agents 224(1)-224(J) may be transferred to the master Al agent 222. The master Al agent 222 may generate the response based on the insight received from each of the set of domain-specific Al agents 224(1)-224(J). In one embodiment, the response can include a summary of the insights received from each of the set of domain-specific Al agents 224(1 )-224(J). In another embodiment, the response can include a consolidation of the insights received from each of the set of domain-specific Al agents 224(1)-224(J).

[0131] In some embodiments, for generating the response, the master Al agent 222 is configured to receive the insight from each of the set of domain-specific Al agents 224(1)- 224(J). Upon receiving the insight from each of the set of domain-specific Al agents 224(1)-P24-016PCT1224(J), the master Al agent 222 may have to check for its validity. Thus, the insight validation module 328 may be configured to facilitate the master Al agent 222 to determine the validity of the insight based, at least in part, on the query and the query execution information.

[0132] In one embodiment, to determine the validity of the insight, the insight validation module 328 may facilitate the master Al agent 222 to access the query execution information from the database 204. In a non-limiting example, the query execution information can include historical queries and their valid responses, activities performed to generate the valid responses, one or more ML models that learn a pattern of generating valid responses to queries, and the like. Thus, it may be understood that the master Al agent 222 is capable of generating a activity execution operation on its own and assigning the activities to different domain-specific Al agents 224 based on matching the intent of the query with the domain of the domain-specific Al agents 224.

[0133] In one embodiment, in response to determining that the insight is valid, the master Al agent 222 may generate the response based on the insight. In another embodiment, in response to determining that the insight is invalid, the master Al agent 222 may select a different domain-specific Al agent for the intent. Herein, the different domain-specific Al agent has a second highest relevancy factor. For example, earlier, for a particular intent, the domain-specific Al agent 224(1) was selected by the master Al agent 222. However, if the insight generated by the domain-specific Al agent 224(1) is invalid or has poor accuracy, then the master Al agent 222 may select another domain-specific Al agent such as the domainspecific Al agent 224(2) having the relevancy factor which is the second highest after that of the domain-specific Al agent 224(1). The process of determining the set of tasks, their respective set of resolutions, and generating the insight is repeated using the another domainspecific Al agent. The master Al agent 222 may again determine the validity of the insight received from the another domain-specific Al agent before generating the response for the query. This process may be repeated until the master Al agent 222 determines that the insights are valid. It is noted that the response is transmitted to the electronic device of the first user 104(1) who asked the query. The response may be displaced to the first user 104(1) on the UI of the electronic device.

[0134] FIG. 3 illustrates a flow diagram 300 depicting a set of organizational processes assigned with a set of domain-specific Al agents, in accordance with an embodiment of the present disclosure. In a non-limiting example, assuming the organization 108 to be a logisticsP24-016PCT1shipment industry 302. In the logistics shipment industry 302, one or more organizational processes 304 include inventory management 304(1), warehousing 304(2), customer service management 304(3), and so on. To improve the performance of the logistics shipment industry 302, these organizational processes 304 may have to be optimized and managed efficiently. These organizational processes 304 may have to be implemented depending upon the requirements of the logistics shipment industry 302.

[0135] In one scenario, the requirements are provided to the organizational processes 304 as a command from a stakeholder such as the first user 104(1). In another scenario, the requirements may be received as input upon implementation of other operations. In yet another scenario, the logistics shipment industry 302 may develop and use one or more domain-specific Al agents 306 for managing each of the one or more organizational processes 304. The domainspecific Al agents 306 are examples of the domain-specific Al agents 224 explained earlier with reference to FIG. 2. In the logistics shipment industry 302, the domain-specific Al agents 306 can include an inventory management agent 306(1) for controlling or the implementation of the inventory management 304(1). The domain-specific Al agents 306 can also include a warehousing agent 306(1) for the warehousing 304(2), a customer service management agent 306(3) for the customer service management 304(3), and so on. Thus, it may be understood that a particular domain-specific Al agent can be used to implement a particular organizational process. These domain-specific Al agents 306 can sense its environment and / or receive commands and / or queries from the first user 104(1), perform one or more activities associated with the respective organizational processes 302, and respond accordingly. However, these domain-specific Al agents 306 may have to be active always for continuously sensing the environment, consuming huge amounts of power and resources. Alternatively, the first user 104(1) may have to be manually activated.

[0136] Therefore, in the present disclosure, the master Al agent 222 is used to control the operation of the domain-specific Al agents 306. In a specific example, the logistics shipment industry 302 has registered with a virtual assistance system 308 which is operatively coupled with the server system 200. Various modules of the server system 200 as explained in FIG. 2 control various operations of the logistics shipment industry through the virtual assistance system 308. It is noted that, during the registration of the logistics shipment industry 302 with the virtual assistance system 308, all the information associated with the logistics shipment industry 302 is accessible to the server system 200. In one scenario, while processing operations of the logistics shipment industry 302, the server system 200 may access thisP24-016PCT1information from different databases associated with the logistics shipment industry 302. In another scenario, the server system may store this information in the database 204 associated with the server system 200, which is accessible for further processing. The information may include different contexts associated with the industry, historical intent information, historical query-response information, agent profile information corresponding to the domain-specific Al agents 306, and the like.

[0137] When a query 310 may be prompted to the virtual assistance system 308 through a prompt 312, the query 310 is transmitted to the master Al agent 222 through the master-slave operation management module 232 of the server system 200. In one scenario, the query 310 may be prompted by the first user 104(1) as shown in FIG. 4. Assuming the first user 104(1) to a logistics and freight optimization manager in the logistics shipment industry 302. The master-slave operation management module 232 may access the predefined context of the query 310 and the historical intent information. This information is made available to the master Al agent 222 for accurately determining the intent associated with query 310. In the logistics shipment industry 302, various non-limiting examples of the query 310 can be as follows:

[0138] Upon understanding the context associated the query 310 which can be any one of these queries, the intent is easily determined by the master Al agent 222. The process of determining the intent of the query 310 has already been explained with reference to FIGS. 1 and 2. To that end, the same is not explained again for the sake of brevity.

[0139] Further, the master-slave operation management module 232 facilitates the master Al agent 222 to select one or more of the domain-specific Al agents 306 based on the intent. The process of selecting one or more of the domain-specific Al agents 306 has already been explained with reference to FIG. 2. To that end, the same is not explained again for theP24-016PCT1sake of brevity. Assuming, the inventory management agent 306(1) and the warehousing agent 306(2) are selected from the domain-specific Al agents 306 shown in FIG. 3. Upon selecting the inventory management agent 306(1) and the warehousing agent 306(2), the master Al agent 222 transmits the query to both the inventory management agent 306(1) and the warehousing agent 306(2).

[0140] Furthermore, the insight generation module 234 facilitates each of the inventory management agent 306(1) and the warehousing agent 306(2) to determine the query category of the query 310 to a task, a goal, or a combination thereof. The insight generation module 234 may further determine a activity execution operation for each of the inventory management agent 306(1) and the warehousing agent 306(2) based on the query category. The type of activities that can be generated for different query categories has been already explained with reference to FIG. 2. To that end, the same is not explained again for the sake of brevity.

[0141] Assuming that the query category determined for a portion of the query 310 at the inventory management agent 306(1) is a task. In an example, the implementation of the task execution operation requires performing a set of sub-tasks. The insight generation module 234 facilitates the inventory management agent 306(1) to generate the set of sub-tasks based on the intent and the availability of tools. As may be understood, the tasks and the sub-tasks are implemented by accessing one or more tools if needed as shown in FIG. 4.

[0142] Now, referring to FIG. 4, a flow diagram depicting a master-slave system architecture used for determining a solution to a problem through a query response generation process is illustrated, in accordance with an embodiment of the present disclosure. In a nonlimiting example, the inventory management agent 306(1) generates the set of sub-tasks as a part of the task execution operation that require access to tools, such as a first tool 402, a second tool 404, and a third tool 406 associated with a first data source 408, a second data source 410, and a third data source 412, respectively. It is noted that inventory management agent 306(1) can have access to any number of tools and any number of data sources without limiting the scope of the approach proposed in the present disclosure.

[0143] The inventory management agent 306(1) may generate a set of resolutions upon implementation of the set of sub-tasks using the above-mentioned tools. Later, the inventory management agent 306(1) may generate an insight from the set of resolutions. Upon generating the insight, it may have to be transmitted to the master Al agent 222. Before sending the insight, the accuracy of the insight is determined. The accuracy of the insight is dependent on theP24-016PCT1accuracy or validity of the set of sub-tasks generated by the inventory management agent 306(1). Thus, the evaluator module 236 determines the validity of the set of sub-tasks using an evaluator 414. The evaluator module 236 transfers the set of sub-tasks and the set of resolutions to the evaluator 414. The evaluator 414 determines the validity of the set of sub-tasks. The process of determining the validity of the set of activities is explained later in the present disclosure. In an example, the evaluator 414 determines that the set of sub-tasks is valid. This validity result is transmitted to the evaluator module 236. Upon determining that the set of subtasks is valid, the insight generated by the inventory management agent 306(1) is transmitted to the master Al agent 222.

[0144] In another non-limiting example, the query category determined for a portion of the query 310 at the warehousing agent 306(2) is a goal. In an example, the implementation of the goal requires steps that are predefined for this particular goal from the database 204. The insight generation module 234 may determine a goal execution operation including a set of tasks to be performed by the warehousing agent 306(2) based on the intent and the predefined goal execution information. The predefined goal execution information may include the steps that are predefined for this particular goal. For the implementation of the goal also, one or more tools may be required as shown in FIG. 4.

[0145] In another non-limiting example, the warehousing agent 306(2) generates the set of tasks that require access to tools, such as a fourth tool 416, a fifth tool 418, and a sixth tool 420 associated with a fourth data source 422, a fifth data source 424, and a sixth data source 426, respectively. It is noted that warehousing agent 306(2) can have access to any number of tools and any number of data sources without limiting the scope of the approach proposed in the present disclosure.

[0146] The warehousing agent 306(2) may receive a set of resolutions upon implementation of the set of tasks using the above-mentioned tools. Later, the warehousing agent 306(2) may generate an insight from the set of resolutions. Upon generating the insight, the accuracy of the insight is determined which is dependent on the accuracy or validity of the set of tasks generated by the warehousing agent 306(2). Thus, the evaluator module 236 determines the validity of the set of tasks using an evaluator 428. The evaluator module 236 transfers the set of tasks and the set of resolutions to the evaluator 428 which determines the validity of the set of tasks. The process of determining the validity of the set of tasks is explained later in the present disclosure. In an example, the evaluator 428 determines that theP24-016PCT1set of tasks is valid. This validity result is transmitted to the evaluator module 236. Upon determining that the set of sub-tasks and / or the set of tasks is valid, the insight generated by the warehousing agent 306(2) is transmitted to the master Al agent 222.

[0147] Upon receiving the insight from the inventory management agent 306(1) and the insight from the warehousing agent 306(2), the insight validation module 238 may facilitate the master Al agent 222 to determine the validity of these insights. The insight validation module 238 may determine the validity of the insights based on the query 318 and the query execution information. In a non-limiting example, the insights from each of the inventory management agent 306(1) and the warehousing agent 306(2) are determined to be valid. Thus, the master Al agent 222 generates a response such as the response 314 (as shown in FIGS. 3 and 4). Now referring back to FIG. 3, it may be understood that the response 314 is displayed to first user 104(1) through the prompt 312.

[0148] FIG. 5 illustrates a flow diagram 500 depicting a process of evaluating a validity of a activity execution operation determined by the set of domain-specific Al agents such as the set of domain-specific Al agents 224(1)-224(J), in accordance with an embodiment of the present disclosure. In a non-limiting example, the inventory management agent 306(1) from FIG. 3 or FIG. 4, generates the insight from the set of resolutions obtained upon implementation of the activity execution operation using the tools 408-412. Further, to determine the accuracy of the insight, the evaluator module 236 determines the validity of the activity execution operation determined by the inventory management agent 306(1). The evaluator module 236 may determine the validity of the activity execution operation using the evaluator 414 (as shown in FIG. 4). The process of evaluating the validity of the activity execution operation starts from step 502.

[0149] At step 502, the activity execution operation determined by the inventory management agent 306(1) is transmitted to a fine-tuned LLM 504. The fine-tuned LLM 504 is an example for the evaluator 414 of FIG. 4. The evaluator module 236 is configured to transmit the activity execution operation from the inventory management agent 306(1) to the fine-tuned LLM 504. It is noted that the fine-tuned LLM 504 is fine-tuned with data that is specific to a domain of the organizational process. In the example considered, the organizational process is the inventory management 304(1). Thus, the fine-tuned LLM 504 may be fine-tuned with the data associated with the inventory management 304(1). The fine-tuning of the fine-tuned LLM 504 may be done during the training period of the fine-tuned LLM 504 based on the predefinedP24-016PCT1training dataset including the domain-specific data described herein. Thus, the data that the fine-tuned LLM 504 is aware of also includes an intent 506 associated with the query 310, steps that need to be performed to resolve the query 310, tools required for this process, and the like. In this example, the intent 506 is related to the inventory management of the logistics shipment industry 302.

[0150] At step 508, the activity execution operation is evaluated by generating the set of fine-tuning parameters for each activity of the activity execution operation. The evaluator module 236 determines the set of fine-tuning parameters based on the intent and the predefined training dataset. More specifically, the evaluator module 236 facilitates the fine-tuned LLM 504 to determine the set of fine-tuning parameters. In a non-limiting example, the set of finetuning parameters includes a rating assigned to each activity along with a comment indicating reasons for the rating and suggestions to improve the rating. The rating can range between 1- 5, 1-10, 1-100, or the like, with 1 indicating the least favorable activities and the maximum limit, i.e., 5, 10, and 100 indicating the most favorable activities. For instance, if the rating is 4 for the rating range of 1-10, then the comment can be asking the inventory management agent 306(1) to re-create better activities as the rating of the activities that are already generated is too low.

[0151] At step 510, the fine-tuned LLM 504 may transmit the set of fine-tuning parameters to the inventory management agent 306(1).

[0152] At step 512, upon receiving the set of fine-tuning parameters, the inventory management agent 306(1) identifies the activity execution operation to be valid or invalid based on the set of fine-tuning parameters and the fine-tuning criteria. In a non-limiting example, the fine-tuning criteria can include a threshold for the rating and identification of phrases within the comment indicating the comment to a positive comment or a negative comment. More specifically, the fine-tuning criteria indicate that the activity execution operation is valid if the rating is at least equal to the threshold and if the comment is a positive comment, otherwise, the activity execution operation is invalid. For instance, in the rating range of 1-5, the threshold can be 3. A rating greater than 3 is considered to be a good rating, hence comment will also be a positive comment. The positive comment can include positive phrases, such as ‘good’, ‘efficient’, or the like. A rating less than 3 is considered to be a bad rating, hence comment will be a negative comment. The negative comment can include negative phrases, such as ‘poor’, ‘incorrect’, ‘make better’, ‘instructions to improve theP24-016PCT1activities generated’, or the like. For example, the comment can include the instruction such as ‘create a task to get a profitability at Y month rather than X month’.

[0153] Therefore, if the set of fine-turning parameters matches the fine-tuning criteria, the inventory management agent 306(1) identifies that the activity execution operation is valid. Alternatively, the inventory management agent 306(1) identified the activity execution operation to be invalid. If the activity execution operation is identified to be invalid, the evaluator module 236 facilitates the inventory management agent 306(1) to generate a new activity execution operation which is also validated using the fine-tuned LLM 504. This process of generating the activity execution operation and the new activity execution operation, and checking their validity is continued until the activity execution operation or the new activity execution operation are identified to be valid. Alternatively, this process may be repeated until a threshold iteration number is met.

[0154] FIG. 6 illustrates a flow diagram of a method 600 for query response generation, in accordance with an embodiment of the present disclosure. The method 600 depicted in the flow diagram may be executed by, for example, the server system 200. The sequence of operations of the method 600 may not be necessarily executed in the same order as they are presented. Further, one or more operations may be grouped and performed in the form of a single step, or one operation may have several sub-steps that may be performed in parallel or in a sequential manner. Operations of the method 600, and combinations of operations in the method 600 may be implemented by, for example, hardware, firmware, a processor, circuitry, and / or a different device associated with the execution of software that includes one or more computer program instructions. The plurality of operations is depicted in the process flow of the method 600. The process flow starts at operation 602.

[0155] At 602, the method 600 includes determining, by a set of domain-specific Artificial Intelligence (Al) agents (e.g., the set of domain-specific Al agents 224(1)-224(J)), a activity execution operation to be performed based, at least in part, on an intent (e.g., the intent 506) associated with a query (e.g., the query 310) and query execution information for the intent 506.

[0156] At 604, the method 600 includes identifying, by the set of domain-specific Al agents 224(1)-224(J), a set of resolutions based, at least in part, on the activity execution operation, each of the set of resolutions being indicative of an outcome of performing a particular activity associated with the activity execution operation.P24-016PCT1

[0157] At 606, the method 600 includes determining, by the set of domain-specific Al agents 224(1 )-224(J), an insight based, at least in part, on the set of resolutions.

[0158] At 608, the method 600 includes generating, by a master Al agent (e.g., the master Al agent 222), a response for the query 310 based, at least in part, on the insight determined by each of the set of domain-specific Al agents 224(1 )-224(J), wherein the master Al agent 222 is communicably coupled to the set of domain-specific Al agents 224(1 )-224(J).

[0159] The disclosed method with reference to FIG. 6, or one or more operations of the server system 200 may be implemented using software including computer-executable instructions stored on one or more computer-readable media (e.g., non-transitory computer- readable media, such as one or more optical media discs, volatile memory components (e.g., DRAM or SRAM), or nonvolatile memory or storage components (e.g., hard drives or solid- state nonvolatile memory components, such as Flash memory components) and executed on a computer (e.g., any suitable computer, such as a laptop computer, netbook, Web book, tablet computing device, smartphone, or other mobile computing devices). Such software may be executed, for example, on a single local computer or in a network environment (e.g., via the Internet, a wide-area network, a local-area network, a remote web-based server, a client-server network (such as a cloud computing network), or other such networks) using one or more network computers.

[0160] 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 are considered to be within the scope of the disclosed technology. Furthermore, any of the software-based embodiments may be uploaded, downloaded, or remotely accessed through a suitable communication means. Such suitable communication means include, for example, the Internet, the World Wide Web (WWW), an intranet, software applications, cable (including fiber optic cable), magnetic communications, electromagnetic communications (including Radion Frequency (RF), microwave, and infrared communications), electronic communications, or other such communication means.

[0161] Although the invention has been described with reference to specific exemplary embodiments, it is noted that various modifications and changes may be made to these embodiments without departing from the broad scope of the invention. For example, the various operations, blocks, etc., described herein may be enabled and operated using hardwareP24-016PCT1circuitry (for example, Complementary Metal Oxide Semiconductor (CMOS) based logic circuitry), firmware, software, and / or any combination of hardware, firmware, and / or software (for example, embodied in a machine-readable medium). For example, the apparatuses and methods may be embodied using transistors, logic gates, and electrical circuits (for example, Application Specific Integrated Circuit (ASIC) circuitry and / or Digital Signal Processor (DSP) circuitry).

[0162] Particularly, the server system 200 and its various components may be enabled using software and / or using transistors, logic gates, and electrical circuits (for example, integrated circuit circuitry such as ASIC circuitry). 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 the processor or the computer to perform one or more operations. A computer-readable medium storing, embodying, or encoded with a computer program, or similar language, may be embodied as a tangible data storage device storing one or more software programs that are configured to cause the processor or computer to perform one or more operations. Such operations may be, for example, any of the steps or operations described herein. In some embodiments, the computer programs may be stored and provided to a computer using any type of non-transitory computer- readable media. Non-transitory computer-readable media includes any type of tangible storage media.

[0163] 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), Compact Disc Read-Only Memory (CD-ROM ), Compact Disc Recordable (CD-R), compact disc rewritable (CD-R / W), Digital Versatile Disc (DVD), and semiconductor memories (such as mask ROM, Programmable ROM (PROM), (erasable PROM), flash memory, Random Access Memory (RAM), etc.). Additionally, a tangible data storage device may be embodied as one or more volatile memory devices, one or more nonvolatile memory devices, and / or a combination of one or more volatile memory devices and non-volatile memory devices. In some embodiments, the computer programs may be provided to a computer using any type of transitory computer-readable media. Examples of transitory computer-readable media include electric signals, optical signals, and electromagnetic waves. Transitory computer-readable media can provide the program to a computer via a wired communication line (e.g., electric wires, and optical fibers) or a wireless communication line.P24-016PCT1

[0164] Various embodiments of the invention, as discussed above, may be practiced with steps and / or operations in a different order, and / or with hardware elements in configurations, which are different than those which are disclosed. Therefore, although the invention has been described based on these exemplary embodiments, it is noted that certain modifications, variations, and alternative constructions may be apparent and well within the scope of the invention.

[0165] Although various exemplary embodiments of the invention are described herein in a language specific to structural features and / or methodological acts, the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as exemplary forms of implementing the claims.P24-016PCT1

Claims

CLAIMS1. A method for query response generation, comprising: determining, by a set of domain-specific Artificial Intelligence (Al) agents, an activity execution operation to be performed based, at least in part, on an intent associated with a query and query execution information for the intent; identifying, by the set of domain-specific Al agents, a set of resolutions based, at least in part, on the activity execution operation, each of the set of resolutions being indicative of an outcome of performing a particular activity associated with the activity execution operation; determining, by the set of domain-specific Al agents, an insight based, at least in part, on the set of resolutions; and generating, by a master Al agent, a response for the query based, at least in part, on the insight determined by each of the set of domain-specific Al agents, wherein the master Al agent is communicably coupled to the set of domain-specific Al agents.

2. The method as claimed in claim 1, further comprising: determining, by the master Al agent, the intent associated with the query based, at least in part, on a predefined context of the query and historical intent information of the master Al agent; and selecting the set of domain-specific Al agents for the intent based, at least in part, on relevancy factors between the set of domain-specific Al agents and the intent, wherein the master Al agent is configured to operate the set of domain-specific Al agents to generate the insight, based at least on the intent.

3. The method as claimed in claim 2, wherein selecting a particular domain-specific Al agent in the set of domain-specific Al agents for the intent comprises: accessing a plurality of agent profiles of a plurality of available domain-specific Al agents; calculating a plurality of relevancy factors between the plurality of agent profiles and the intent; and selecting the domain-specific Al agent associated with an agent profile that has the highest relevancy factor from the plurality of relevancy factors.P24-016PCT14. The method as claimed in claim 1 , wherein determining the activity execution operation comprises: identifying a query category for the intent based, at least in part, on predefined category information, the query category comprising one of a task, a goal, or a combination thereof.

5. The method as claimed in claim 4, wherein determining the activity execution operation further comprises: upon identifying the query category to be the task, generating, dynamically, a task execution operation comprising a set of sub-tasks to be performed for generating the insight based, at least in part, on the intent and task-related information; and assigning the task execution operation to the activity execution operation to be performed.

6. The method as claimed in claim 4, wherein generating the activity execution operation further comprises: upon identifying the query category to be the goal, determining a goal execution operation comprising a set of tasks to be performed for determining the insight based, at least in part, on the intent and predefined goal execution information; and assigning the goal execution operation to the activity execution operation to be performed.

7. The method as claimed in claim 4, wherein generating the activity execution operation further comprises: upon identifying the query category to be a combination of the task and the goal, segregating the query into at least two sub-queries comprising a first sub-query indicative of the task and a second sub-query indicative of the goal; generating and assigning, dynamically a task execution operation to the activity execution operation for generating the insight for the first sub-query; and determining and assigning a goal execution operation to the activity execution operation for generating the insight for the second sub -query.P24-016PCT18. The method as claimed in claim 1, wherein generating the insight further comprises: determining a validity of the activity execution operation based, at least in part, on the activity execution operation, the set of resolutions, and a Large Language Model (LLM).

9. The method as claimed in claim 8, wherein determining the validity of the activity execution operation comprises: accessing the activity execution operation and the set of resolutions; determining, by the LLM, a set of fine-tuning parameters for each activity of the activity execution operation based, at least in part, on the intent and a predefined training dataset of the LLM; and transferring, by the LLM, the set of fine-tuning parameters to the domain-specific Al agent based at least on the intent.

10. The method as claimed in claim 9, wherein determining the validity of the activity execution operation further comprises: receiving, by the domain-specific Al agent, the set of fine-tuning parameters from the LLM; and performing one of: identifying the activity execution operation to be valid when the set of fine-tuning parameters matches with fine-tuning criteria; or identifying the activity execution operation to be invalid when the set of fine-tuning parameters deviates from the fine-tuning criteria.

11. The method as claimed in claim 10, wherein upon identifying that the activity execution operation is invalid, performing: generating a new activity execution operation based, at least in part, on the intent and the set of fine-tuning parameters; identifying a new set of resolutions for the new activity execution operation based, at least in part, on performing the new activity execution operation; and determining a validity of the new activity execution operation based, at least in part, on the new activity execution operation, the new set of resolutions, and the LLM.P24-016PCT112. The method as claimed in claim 10, wherein generating the response comprises: receiving, by the master Al agent, the insight from each of the set of domain-specificAl agents; determining, by the master Al agent, a validity of the insight based, at least in part, on the query and the query execution information; and in response to determining that the insight from each of the set of the domain-specific Al agents is valid, generating, by the master Al agent, the response, based at least on the insight.

13. The method as claimed in claim 10, wherein in response to determining that the insight from each of the set of domain-specific Al agents is invalid, the method further comprises selecting a different domain-specific Al agent for the intent, wherein the different domain-specific Al agent has a second highest relevancy factor.

14. A server system, comprising: a communication interface; a memory configured to store instructions; and a processor in communication with the communication interface and the memory, the processor configured to execute the instructions stored in the memory and thereby cause the server system to perform at least in part to: determine, by a set of domain-specific Artificial Intelligence (Al) agents, a activity execution operation to be performed based, at least in part, on an intent associated with a query and query execution information for the intent; identify, by the set of domain-specific Al agents, a set of resolutions based, at least in part, on the activity execution operation, each of the set of resolutions being indicative of an outcome of performing a particular activity associated with the activity execution operation; determine, by the set of domain-specific Al agents, an insight based, at least in part, on the set of resolutions; and generate, by a master Al agent, a response for the query based, at least in part, on the insight determined by each of the set of domain-specific Al agents, wherein the master Al agent is communicably coupled to the set of domain-specific Al agents.P24-016PCT115. The server system as claimed in claim 14, wherein the server system is further caused, at least in part, to: determine, by the master Al agent, the intent associated with the query based, at least in part, on a predefined context of the query and historical intent information of the master Al agent; and select the set of domain-specific Al agents for the intent based, at least in part, on relevancy factors between the set of domain-specific Al agents and the intent, wherein the master Al agent is configured to operate the set of domain-specific Al agents to generate the insight, based at least on the intent.

16. The server system as claimed in claim 15, wherein to select a particular domain-specific Al agent in the set of domain-specific Al agents for the intent, the server system is caused, at least in part, to: access a plurality of agent profiles of a plurality of available domain-specific Al agents; calculate a plurality of relevancy factors between the plurality of agent profiles and the intent; and select the domain-specific Al agent associated with an agent profile that has the highest relevancy factor from the plurality of relevancy factors.

17. The server system as claimed in claim 14, wherein to generate the insight, the server system is caused, at least in part, to: determine a validity of the activity execution operation based, at least in part, on the activity execution operation, the set of resolutions, and a Large Language Model (LLM).

18. The server system as claimed in claim 14, wherein to generate the response, the server system is caused, at least in part, to: receive, by the master Al agent, the insight from each of the set of domain-specific Al agents; determine, by the master Al agent, a validity of the insight based, at least in part, on the query and the query execution information; and in response to determining that the insight from each of the set of the domain-specificP24-016PCT1Al agents is valid, generate, by the master Al agent, the response, based at least on the insight.

19. The server system as claimed in claim 18, wherein in response to determining that the insight from each of the set of domain-specific Al agents is invalid, the server system is further caused, at least in part, to: select a different domain-specific Al agent for the intent, wherein the different domainspecific Al agent has a second highest relevancy.

20. A non-transitory computer-readable storage medium comprising computer-executable instructions that, when executed by at least a processor of a server system, cause the server system to perform a method comprising: determining, by a set of domain-specific Artificial Intelligence (Al) agents, a activity execution operation to be performed based, at least in part, on an intent associated with a query and query execution information for the intent; identifying, by the set of domain-specific Al agents, a set of resolutions based, at least in part, on the activity execution operation, each of the set of resolutions being indicative of an outcome of performing a particular activity associated with the activity execution operation; determining, by the set of domain-specific Al agents, an insight based, at least in part, on the set of resolutions; and generating, by a master Al agent, a response for the query based, at least in part, on the insight determined by each of the set of domain-specific Al agents, wherein the master Al agent is communicably coupled to the set of domain-specific Al agents.P24-016PCT1

Images