A NEW READY-MADE ARTIFICIAL INTELLIGENCE AGENT PLATFORM

Abstract

The invention relates to an artificial intelligence agent platform system that enables artificial intelligence agents to perform business processes carried out in many units such as technology, marketing, sales, finance, purchasing, law, regulation, human resources, customer (803) relations, which constitute the organization of telecommunication operators (804) and Service Providers in the telecommunications sector, either in place of people or as a support to people.

Description

1 TARIFF A NEW READY-MADE ARTIFICIAL INTELLIGENCE AGENT PLATFORM Technical Area The invention is intended for telecommunications operators and service providers in the telecommunications sector. The providers' organization comprises technology, marketing, sales, finance, purchasing, 5 performed in many departments such as law, regulation, human resources, and customer relations using artificial intelligence agents to replace or support people in business processes. It relates to the artificial intelligence agent platform that enables this. State of the Art Fixed and mobile telecommunications operators and service providers in the telecommunications sector 10 providers offer services such as internet, voice, and IPTV to individual customers, and services to corporate customers; security, cloud services, point-to-point data transfer, data transfer between multiple points They offer services such as transportation. Telecommunications Operators and Service Providers, Jobs created according to the responsibilities and duties of the units within the organizations. It is managed through processes. These business processes include technology, marketing, sales, finance, purchasing, 15 It can be applied in many areas such as law, human resources, and customer relations. Currently... Some business processes are pre-designed, implemented, and integrated with each other. with software applications that work, some of which are initiative-based and human-centered, even without a digital environment. It is carried out through skill. Developing and integrating software applications for each new business process 20 It takes a lot of time and is very costly. This also means that market opportunities require new products and services. This can lead to them leaving during their development. These applications employ human workers. It is used by [the employer]. All operations are susceptible to human error. New hires Employees, however, do not know how to use the system. The adaptation of new employees... It may take a long time. 25 In business processes, transferring every detail of the process to rule-based software applications. Due to complexity and technical impossibility, it is often not possible. In-house work. revisions to existing processes by the units designing the process in changing market conditions It takes a long time for software development units to meet their requests and new demands. It takes time, and generally all rule-based software needs to undergo fundamental changes. 30 2 This is necessary. This situation results in significant time and budget costs, even for small requests. Why does that happen? Business Support Systems and Operations Support Systems, which do not have a digital environment, Small and medium-sized telecommunications companies that manage their business processes through human ingenuity, depending on the initiative. Numerous human-related errors, problems, and 5 issues exist among operators and service providers. Problems arise. Additionally, long queues for jobs are due to a shortage of human resources. It's been a long time. Creating job-specific employment opportunities in these companies is also quite difficult, Expertise cannot be developed. Therefore, disruptions and errors occur in technical and commercial processes, and Unsolved problems accumulate. Existing agent platforms lack autonomous evaluation and feedback mechanisms, or 10 It lacks or has partial feedback mechanisms that are not fully autonomous. The system measures itself and generates quantitative metrics, which in turn analyzes these metrics using human feedback. with feedback mechanisms created and quantified through their notifications They lack integrated structures. In conclusion, due to the negative aspects described above and the current solutions regarding issue 15 Due to its shortcomings, an improvement is needed in the relevant technical field. It has been made. Purpose of the Invention The invention was created by drawing inspiration from existing situations and overcoming the aforementioned drawbacks. It aims to solve. 20 The main purpose of the invention is to enable telecommunication operators and those involved in the telecommunications sector. The technology, marketing, sales, finance, and purchasing components that make up the organization of Service Providers. It is used in many departments such as procurement, law, regulation, human resources, and customer relations. artificial intelligence in place of or as a support to humans in the introduced business processes The goal is to ensure that its agents carry it out. 25 The platform that is the subject of the invention can be used in sectors other than telecommunications; energy, manufacturing, retail, in many vertical sectors such as entertainment, e-commerce, logistics, agriculture, healthcare, education, transportation, and security. It is available for use. 3 Through this invention, businesses can use this platform to acquire licensed agents through the platform. It enables units to perform specialized work processes without human intervention. The agents on the platform are made available for the company's use. Invention reduces the costs of developing and implementing new products, services, and features. The purpose of the invention is to implement the requests made by units that improve business processes. 5 The goal is to shorten the time frame. In inventions, revisions can range from months to hours or even minutes. It can descend to that level. In this invention, layers of security can be integrated into the process. The platform itself is natural. Because it is an AI application, new AI services can be added to the system much faster. It is being integrated. 10 The invention facilitates architectural changes on the platform that provides the solution. The invention relates to protocols such as A2A (Agent2Agent) and ACP (Agent Control Protocol). It ensures its implementation. In the invention, the platform is described as a multi-agent platform. It is located. 15 The platform that is the subject of the invention has autonomous evaluation and feedback mechanisms. It has the capability to create quantitative evaluation metrics by self-measuring. feedback metrics are created and quantified from human feedback. It has structures that integrate with its mechanisms. This allows it to be used by artificial intelligence agents. Every work process being implemented is constantly improving, making the system's performance even higher. It is coming. The structural and characteristic features and all the advantages of the invention are given in the figures below. This becomes clearer thanks to the detailed explanation written with references to these figures. This will be understood, and therefore the evaluation should also take these forms and detailed explanations into consideration. It needs to be done by taking precautions. 25 The structural and characteristic features and all the advantages of the invention are given in the figures below. This becomes clearer thanks to the detailed explanation written with references to these figures. This will be understood, and therefore the evaluation should also take these forms and detailed explanations into consideration. It needs to be done by taking precautions. 4 Figures that will help understand the invention. Figure 1 shows a block diagram view of the platform that is the subject of the invention. Explanation of Part References 101. Platform management module 102. Intent Detection Module 5 103. Event Trigger Module 104. Business Process Flow Module 105. Timer module 106. Web research agent module 107. Platform channel module 10 201. Input security layer 202. API gateway 203. Identity and access management 301. Large language model 302. Planning agent 15 303. Evaluation agents and feedback module 304. Predefined mission agents 305. Vehicle caller 401. Adapter 402. Business support systems 20 403. Operation support systems Connector 404. 405. External software applications 406. Fixed and mobile network devices 407. Other artificial intelligence application tools 25 408. Databases 409. External cloud services 410. Servers 501. Intelligent operation module 502. Structured data 30 503. Unstructured data 504. Knowledge base 505. Synthetic data 506. Telecommunications memory 601. Business Process Pool Module 701. Data lake and repository 702. Monitoring & Recording 703. Counter module 5 704. Pricing module 801. Admin 802. Developer Customer 803. 804. Operator 10 Detailed Description of the Invention This detailed explanation only serves to better understand the subject of the invention. This is explained as follows. The invention applies to operators (804) and service providers operating in the telecommunications sector. providers, technology, marketing, sales, finance, procurement, law, regulation, human resources human resources and customer (803) relations are processes carried out in different business units. without human intervention or with human support, through artificial intelligence agents It is an artificial intelligence agent platform system that brings together all aspects of the platform, and the system administrator has complete control over the platform. modules, agents, knowledge base (504), data sources, business process pool and tools 20 A platform that enables the configuration, management, and monitoring of context protocols. Management module (101) analyzes incoming triggers, selects the appropriate business process, and assigns the task breaking down the process into subtasks, assigning subtasks, feedback from process evaluation agents Planning agent (302) that constantly revises according to notifications, user or system origin Dual 25 that transmits triggers to the platform via voice, text, or data-based protocols. Platform channel module (107) that provides one-way communication, user or artificial intelligence It analyzes natural language input provided by the user and deduces user intent from the input. intent detection module that determines and transmits to the planning agent (302) in a standardized format (102), monitors technical, operational or network-based events in the system, changes or Event trigger module (103), which triggers the planning agent (302) when it detects an anomaly, 30 Maintaining the sequence and conditional relationships between multiple business processes, specific events or intentions. when it happens, the relevant business process is implemented by the planning agents (302) Business process flow module (104) that provides specific business processes by admin (801) enable it to start automatically at defined times or periods 6 Timer module (105) which triggers the planning agent (302), external data sources or monitors the web environment, and when a specific development or change is detected, the platform is activated. triggering web search agent module (106), in telecommunication business processes performing repetitive sub-tasks without human intervention predefined task agents (304), predefined task agents (304) need 5 the system it hears about, enabling it to interact with network devices, databases, or external services; Tools that call contextually necessary tools via API or custom protocols. by caller (305), planning agents (302) and predefined task agents (304) the results of the processes performed in terms of performance, time, resource usage and user Evaluation agents and feedback module 10 that assess using criteria such as satisfaction (303) divides the tasks of planning agents (302) in the business process into subtasks. when prioritizing, ordering and using predefined task agents (304) business process-related references used to enable accurate decision-making during matching. Business process pool module (601) that implements the advice, suggestions and directives of the admin (801) It includes. Assessment agents and feedback module (303), reinforcement learning 15 using algorithms to determine the decision policies of planning agents (302) It is updating. The platform management module (101) is where the admin (801) accesses the system and configures system settings. and is the module that manages the system. This is necessary for the system to perform its tasks completely. Necessary information updates are made through this module. 20 Admin (801) uses the platform management module (101) for the following purposes;  intent detection module (102), event trigger module (103), business process flow module (104), timer module (105), web search agent module (106) when he will configure it,  When it will manage the intelligent operation module (501), 25  when it will manage the structured data (502),  when it will manage unstructured data (503),  when he will manage the knowledge base (504),  when it will manage synthetic data (505),  When managing the business processes (601) in the business process pool module, 30  when it will manage predefined task agents (304),  the vehicle caller (305); that is, the context definitions (model context protocol / model when it will manage the context protocol, 7  When the adapter (401) will manage the system. Intent detection module (102), from admin (801), from customers (803), from operator (804) or new input from natural language of any human or artificial intelligence with a role in it; subscription, subscription transfer, speed upgrade, package upgrade, inventory forecasting, purchasing, Intent in requests such as capacity determination or complaints such as malfunctions or billing disputes 5 He informs the planning agent (302) by understanding and detailing it in a specific format. The planning agent (302) selects the most suitable module (601) in the business process pool in line with the intent. He wants to implement the business process. The event trigger module (103) detects the change when there is a change in the whole system, detailing in a specific format, triggering planning agents (302) and change 10 It is a module that facilitates detailed information on the subject. For example; smart operations. When module (501) detects an anomaly in the system, event trigger module (103) It detects the change and uses the information it receives from the intelligent operation module (501) to process this information; what the anomaly was, where it occurred, when it occurred, how long it lasted, its effects its magnitude, and if a problem is foreseen, how the problem can be prevented, etc., in a format like this. 15 (302) directly or indirectly. In this regard, planning agents (302) business process pool It wants to implement the most suitable business process in module (601). Business process flow module (104), via the admin's (801) platform management module (101) It maintains the business process flows it has created; when conditions and requirements arise, this work... Planning 20 to be implemented one-to-one by the planning agents (302) in the sequence of flows. It is the module that triggers the agents (302) and provides the flow information to the planning agents (302). This module triggers planning agents (302) to serially or parallelize appropriate business processes. It enables implementation. Flows defined on the business process flow module (104): business Not detailed steps to be taken within the processes, but multiple solutions to be found. 25 in situations where the business process needs to be implemented, between the business processes themselves determines the order of implementation. The business process flow module determines the planning agents (302) the necessary information for triggering; from the intent detection module (102), event trigger from module (103), timer module (105), web search agent module (106) and can obtain it from any defined source inside and outside the system. Timer module (105): admin (801), via platform management module (101) 30 He manages the specific business processes he defines to ensure they are implemented at the desired times. It is a module. It ensures the precise execution of certain business processes at predetermined times. It is necessary. Admin (801) sets the time for the execution of the work process, timer 8 It does this by matching the time information with the work process on module (105). Timer Planning agents (302) for each business process whose implementation time is due for module (105) It triggers. For example, every night at a certain hour and minute, someone whose birthday is the next day. will segment customers (803), identify the most suitable gift package for these segments 5 of the business process which will define the gift wrapping for customers (803) to use. All timing settings for implementation are on the timer module (105) The timer module (105) is used to schedule the agents for each business process in turn. (302) triggers directly or via the business process flow module (104). Web research agent module (106): admin's (801) platform management module (101) He defined it through, constantly searching for a new development in the outside world, and when development occurs... It is the module that triggers the system when it arrives. Platform channel module (107): customers (803), admin (801), operator (804), Two-way communication via the platform; text-based communication, speech-based communication, and through special protocols. This module will enable communication via message. This module also provides input to the system. will provide; any software application, inside or outside the platform, 15 any hardware application, inside or outside the platform, inputs into the system any artificial intelligence application, inside or outside the platform that will provide this, any written, verbal, or message-based communication that establishes contact between the platform and the user. It refers to communication channels. Communication channels are based on spoken language, written communication, Application programming interfaces can be based on data flow, as well as all custom 20 It includes access to the platform by all types of communication tools via protocols. Platform channel module (107) integrates the following:  an artificial intelligence call center assistant that communicates two-way with the customer (803),  A human call center assistant who communicates two-way with the customer (803) and its system, 25  providing two-way communication with the customer (803) and through the platform and services an integrated mobile application (including data types such as voice, text, image, event notification, etc.) includes.),  providing two-way communication with the customer (803) and through the platform and services an integrated web application (supporting 30 data types including voice, text, image, event notification, etc.) includes.),  Two-way communication with the customer (803) and integrated mail and SMS with the platform etc. special services, 9  Application software services that are integrated with the platform and communicate bidirectionally. (REST, GraphQL, gRPC, WebSocket, SOAP, WebRTC, QUIC, etc.)  IoT (Internet of Things) that is integrated with the platform and communicates bidirectionally. (Internet) or services between closed-system hardware (MQTT, CoAP, etc.),  System 5 that requires platform integration and near real-time data flow messages (Kafka, RabbitMQ, etc.) It includes. Each component in the platform channel module (107) is also in the agent architecture. It is a tool. The platform channel module (107) included in the invention connects all types of communication devices, communication It can include the protocol and the communication system. 10 Input security layer (201): Security, confidentiality and authorization of all requests coming to the system. It is the filtering layer in terms of data. It blocks malicious or unauthorized requests; sensitive It filters out content that may violate company policies, such as personal data, during the entry process. It prevents attacks on the system. API gateway (202): Manages all API calls between the system and external applications. It is the software and the hardware on which the software runs. Ensuring that requests are delivered to the correct service. It enables routing, format conversions, and load balancing. It also... It acts as an external service integration point and ensures the secure and controlled flow of API calls. does. Identity and access management (203): Authenticates users accessing the system. It is the unit that manages role-based authorization. Which user or system component, It controls which data and operations it can access. Big language model (301): Trained in telecommunications operations and processes (and / or (Finely tuned with relevant information) artificial intelligence language model. Both user experience in natural language It is used both to understand their needs and to interpret data and generate recommendations. Large language 25 model (301), planning agents (302) within the platform, evaluation and feedback It provides information to its agents, predefined task agents (304) and to the user when needed. It generates the responses to be given. Planning agents (302): Appropriate action according to the incoming request, complaint, notification or trigger They are intelligent agents who prepare the plan. From the intent detection module (102), event trigger 30 from module (103), from business process flow module (104), timer module (105), interprets the triggers from the web research agent module (106) and the task It creates the most suitable business processes from those included in the business process pool module (601). It identifies what is / are happening by referencing the task and provides suggestions, recommendations, and advice in these business processes. Taking the instructions into account, he / she divides the task into sub-tasks, and arranges the order of the sub-tasks. It determines. It matches each decomposed subtask with the predefined task agents (304) 5 and assigning sub-tasks to predefined task agents (304) to carry them out It does this. It continuously monitors the progress of each sub-task, and uses evaluation agents and According to the feedback from the feedback module (303), each stage is redone It makes revisions by making updates. Planning agents (302); 10  When any intention is detected from the intention detection module (102) and / or  Which event occurred, which status changed, which from the event trigger module (103) when a change is detected as a result of observation and / or  Determine which business process is next in line in the workflow from the business process flow module (104) when and / or 15  The operating time is set or adaptive from the timer module (105). when it is determined which business process triggered the time and / or  Which research activity result in the web research agent module (106) when reached, is triggered. Planning agents (302) meaningful 20 in the format of the triggering notification Tasks are created by taking this information into consideration. Planning agents (302) determine which work in the business process pool module (601) this task belongs to. It determines that it will be fulfilled by implementing the process. Planning agents (302) divide the task into sub-tasks, taking into account the business process. Planning agents (302) have which predefined task 25 to perform these sub-tasks. It determines that its agents (304) will be operated, sub-tasks are predefined tasks He delegates to his agents (304). Planning agents (302), evaluation agents and feedback from the module (303) Based on feedback, the task it infers after triggering is the task matched with the task. the process, the sub-tasks created by breaking down the task, the sub-tasks to which they are assigned previously 30 You can revise the defined task agents (304) in whole or in part. 11 Planning agents (302), developers (802) are a component of the platform that is the subject of the invention. It is pre-created at the code level. Planning agents (302) operate on a plug-and-play principle. It works. These agents are not created via a drag-and-drop method through the platform. Developers should call (802) when changes are needed in the structure and connection points of the agents. They do this by developing and implementing the change. 5 Evaluation agents and feedback module (303): Planning agents (302) fulfill their responsibilities and pre-defined task agents (304) After completing the sub-tasks, analyze the result and suggest improvements. It is a component that monitors and evaluates the results of the processes carried out and, if necessary, incorporates them into the plan. They are intervening agents. Each mission result and system output is 10 times the expected target. It evaluates the situation and suggests new actions (repetition, alternative solutions, etc.) if necessary. It also provides feedback on system performance, leading to model and process improvement. It contributes. In the platform that is the subject of the invention, the "input" for the business processes in the business process pool module (601) The 15 in the chain "trigger → process → task → subtask → task agent → tool" Mappings are defined once in advance by admin (801). These definitions are used for planning. Select a business process for each trigger by the agents (302) and associated with the business process The task is to break it down into sub-tasks and match it with appropriate task agents and vehicles. It is accepted as a reference. Enabling a feedback mechanism will facilitate these matching processes. Increasing its effectiveness and adaptability is critical. For this, the platform has two 20 Different feedback sources are used together: 1. Evaluation Agents – These agents monitor the progress, duration, and resources of each sub-task. continuously through performance metrics such as usage and user satisfaction evaluates the results of the task and subtask planning of the planning agents (302). It transmits this as metric-based feedback to the system for correction. 25 2. Reinforcement Learning from Human Feedback conducted by Admin (801) (RLHF) / Reinforcement Learning from Human Feedback, the learning process A reward that aligns with human preferences, rather than adhering to a classic reward function. It is implemented with this model. In this approach, a reward model is used using human feedback. is created and the model, through reinforcement learning, maximizes this reward signal. 30 It makes fine adjustments. Modern IGBPÖ processes consist of multiple stages: primarily major The language model (301) is fine-tuned in a controlled manner with human-labeled data; then human 12 Users provide data for training the reward model by ranking the model's outputs; final At this stage, a reinforcement learning algorithm such as Proximal Policy Optimization (PPO), The model optimizes itself to maximize this reward signal. The role of evaluation agents is to assess the time, cost, or failure of the business processes being performed. It monitors measurable outputs such as the rate and provides quantitative feedback to planning agents (302). 5 Agents reschedule failed subtasks, predefined task agents (304) or it leads to revisions such as changing the tools to be used. Thus, the system processes the data. It has a learning cycle based on... The role of the IGBPÖ (International Union for Political, Economic and Social Research) is to incorporate preferences into the system that are more abstract and difficult to express with quantitative criteria. It pays. Human operator (804) or admin (801) can match different results 10 It compares and allows users to choose which one is better through a user interface. It reports; a reward model is trained with this preference data. The reward model enables planning agents to... (302), input trigger → business process → task → subtask → task agent → tool selection "What is more suitable or beneficial for humans?" for the steps that emerge in the chain It generates a numerical signal regarding this. This signal is used for planning with algorithms such as PPO. 15 updates the policy parameters of its agents (302) and thus the matching policy It ensures alignment with human preferences. Predefined task agents (304): Certain items frequently needed in telecommunications business processes. The admin (801) is to perform subordinate tasks without the need for human intervention. These are specialized agents developed and deployed prior to the platform's use. For example, 20 Creating a subscription, opening a service request, configuring a device, changing Wi-Fi settings They can autonomously perform specific tasks, such as optimizing according to conditions. They carry out specific sub-tasks mapped by planning agents (302). Defined task agents (304) perform the function of the relevant vehicle via the vehicle caller (305). It runs. In the application, predefined task agents (304), developers (802) need 25 They develop it accordingly and make it available on the platform. Examples of some predefined task agents (304) are as follows. Please refer to this example. all business processes and tasks that can be implemented with artificial intelligence agents, excluding agents. It includes sub-tasks.  Subscriber Assistant: Receives new subscription applications, verifies them, and determines the appropriate package. recommends.  Identity Authenticator: Performs identity, document, e-signature, and facial recognition checks. 13  Customer Profiler: Performs user segmentation and suggests campaigns.  Cancellation Manager: Manages the subscription cancellation process and creates loyalty offers.  Setup Orchestrator: Provisioning devices such as modems, ONTs, STBs, and IP phones. does.  Service Activator: Activates services such as internet, IPTV, and VoIP. 5  Service Availability Auditor: Conducts infrastructure and service suitability checks.  Fault Hunter: Detects faults, performs self-healing, and provides solutions. provides.  Performance Observer: Network Service Quality and Customer Experience Quality It tracks (QoS / QoE) metrics. 10  Cause Analyst: Identifies the cause of the failure and performs correlation analysis.  Prediction Expert: Predicts device malfunctions in advance.  Stock Tracker: Monitors inventory such as devices, cards, and modems.  Dispatch Planner: Optimizes installation teams in terms of routes and inventory.  Delivery Observer: Tracks the courier / cargo process. 15  The virtual agent serves as a chatbot.  Complaint Referrer: Classifies complaints and forwards them to the relevant unit.  Satisfaction Analyzer: Analyzes customer (803) comments.  Campaign Engineer: Creates personalized campaigns and offers.  Revenue Optimizer: Optimizes package pricing. 20  Invoice Auditor: Identifies invoice discrepancies.  Collection Coordinator: Tracks overdue payments and suggests payment plans.  Fraud Detective: Detects suspicious usage and fraud.  Base Controller: Manages RRU / BBU / DU, adjusts frequency and capacity settings. It arranges. 25  Coverage Monitor: Analyzes mobile coverage areas and identifies weak areas.  Load Balancer: Optimizes traffic load between cells.  GDPR Protector: Checks the legality of data processing operations.  Threat Monitor: Detects cyberattacks and unusual behavior.  Listening Manager: Manages legal listening requests. 30 Vehicle caller (305): Interacts with predefined mission agents (304), Programmable application interfaces (APIs) that provide the necessary tool functions depending on the context. It is the component that calls through. It will be implemented to carry out telecommunications business processes. 14 All actions in all sub-tasks are run via the vehicle caller (305) It is taken in the vehicle. Business support systems (402) (BSS-Business Support) are used in the work to be done. System) and operation support systems (403) (OSS-Operation Support System) all software applications, network devices, databases (408), external web at the layer contextual relationship tool caller (305) for applications and systems 5 is being created. Vehicle caller (305) MCP (Model Context Protocol / Model Context It can use standardized protocols (such as the protocol) or specific protocols. Adapter (401): Protocol between agent architecture and vehicles on the platform that is the subject of the invention. These are interface components that enable transformations and internal integration. Predefined task 10 business support systems (402) and operations via agents (304), vehicle caller (305) systems such as support systems (403) communicate in their own protocols and formats This makes it possible. For example, to add a billing system or network management software. The calls are standardized thanks to the relevant adapter (401). For example, the communication protocols of some network devices are as follows: Adapter (401), Programmable tool called by caller (305) in REST API JSON format It converts application interfaces to the following network management protocols.  SNMP (Simple Network Management Protocol): Device monitoring, alarm collection.  TR-069 / CWMP: Remote management of end devices such as modems / routers 20  Netconf / YANG: Programmable management of modern network devices  CLI (Command Line Interface): Direct access to routers / switches (SSH, Telnet)  REST / SOAP API: Integration of BSS / OSS systems and AI agents  xDSL / GPON / DOCSIS: Interfaces of physical link technologies  AAA (RADIUS, Diameter): Authentication and authorization systems 25  Syslog: Collecting logs from devices  MQTT / Kafka / AMQP: Telemetry and real-time messaging infrastructures  TL1: Remote management of certain network devices The protocols converted by the adapter (401) on the platform that is the subject of the invention are 30 Not limited to the above. Every new and old protocol can be added to the adapter (401). Business support systems (402): Telecommunication operator (804), service provider or any for use in a company's business processes; either sourced externally or developed in-house. or a platform that is fully compatible and integrated with the platform it procures from its platform supplier; Existing software that handles business operations such as customer (803) management, billing, CRM. These are applications. Business support systems (402) are applications of businesses that are users of the platform. These could be platform-independent purchased software applications that run locally. such as, integrated into the platform by the product supplier to platform users The service provided may also be software as a service (SaaS) via the cloud. 5 Business support systems (402) software that operators can purchase independently After purchasing the platform, the operator integrates the components one by one into the platform later on. According to the request, cloud-based business support systems (402) software service is also the subject of the meeting. It comes integrated with the platform and is included with the purchase. Cloud-based business With the support systems (402) software, the Operator can directly access all 10 without doing any development work. It can utilize processes through automation. The platform within the scope of the invention communicates with these systems via an adapter (401) to perform work It carries out its processes. Some business support systems (402) are as follows;  Customer Relationship Management: Customer (803) data, interaction management 15  Billing Systems: Billing, campaigns, payment tracking  Online Charging System: Instant charging for prepayment / postpayment systems.  Order Management System: Subscription order and service process flow management  Campaign Management: Promotion, discount and customer (803) targeting tools  Self-Service Portal: Customer (803) panel, chatbot, digital channel management 20  Survey System: Survey collection and analysis  Customer Support System: Used to manage support requests. Each component within the scope of business support systems (402) is one in the agent architecture. It is a tool. The platform that is the subject of the invention has many features beyond what is written in my application document. There are a number of different business support system (402) components. 25 Operation support systems (403): Telecommunication operator (804), service provider or for use in the business processes of any company; whether procured from external sources or manufactured by itself. fully compatible with the platform it developed or procured from the platform supplier, and Integrated systems: network management, service activation, device configuration, alarm. existing software that runs technical operations such as processes and performance monitoring 30 or servers (410). For example, NMS (Network Management System) network devices, 16 ACS (Auto Configuration Server) monitors / manages remote modem devices. These are systems. Alarm management systems, performance management systems, configuration management systems, inventory systems including auto-discovery and correction systems, and similar systems. It is included in the operational support systems (403). The platform is an adapter (401) to these systems. It performs technical tasks by accessing through it. Operation support systems (403), 5 platform-independent buyers working locally for businesses that are users of the platform These could be software applications acquired, or product suppliers integrated into the platform. Software services provided to platform users via the cloud (Software) (as a Service / SaaS) Operation support systems (403) software that operators can purchase independently 10 After the operator purchases the platform, the components are subsequently integrated into the platform one by one. Cloud-based operational support systems (403) software according to the operator's request The service is also integrated with the platform that is the subject of my invention, included in the purchase. It is coming. Cloud-based operational support systems (403) software with the Operator It can directly use automation for all processes without doing any development work. 15 Some operational support system (403) components are as follows.  Inventory Management Application: Manages network and equipment inventory.  Provisioning and Activation System: Service definition and network devices It manages its configuration.  Workforce Management System: Plans field teams and ensures teams operate on a 20-person basis. It manages the appointment calendar.  Network Management System: General management of network components.  Element Management System: Manages devices belonging to specific suppliers.  Service Assurance Tools: Performance, fault, and alarm analysis  Topology Discovery Tools: Automatic network mapping and correlation 25 Each component agent architecture included within the scope of operational support systems (403) They are tools. Apart from what is written in my application document, the platform that is the subject of the invention... There are numerous different operational support system (403) components. Connector (404): Integration with external third-party software, databases, or devices. These are the general interface components that provide. For example, the operator's (804) external software 30 applications (405), fixed and mobile network devices (406), Other Artificial Intelligence Applications The connection between the tools and the databases (408) is established via the appropriate connector (404). 17 External software applications (405): Located within the telecommunications ecosystem, but within the scope of the business support systems (402) and operational support systems (403) of the enterprise These are external software applications that are not part of the company's operations. For example, an external customer support platform, a third-party application. a side analysis tool, notification systems, map applications and similar software and These systems can be evaluated within this scope. 5 Each component within the scope of external software applications (405) in the agent architecture They are tools. Fixed and mobile network devices (406): In the infrastructure where the telecommunications operator (804) provides services located in a network such as routers, switches, transmission systems, base stations, modems, and access points. These are all hardware and software network devices and equipment. Predefined task 10 agents (304), directly or through vehicle caller (305) or operational support systems (403) operations such as setting up, configuring or testing on these devices He can. Managed fixed and mobile systems are used to implement technical business processes. Examples of network devices are as follows: 15  OLT (Optical Line Terminal): Located at the heart of the fiber infrastructure; connected to ONTs. provides service.  ONT (Optical Network Terminal): This is FTTH end-user equipment used in the home. is established.  DSLAM (DSL Access Multiplexer): An access point in DSL-based connections. 20  Modem / VDSL / ADSL CPE: Allows users to connect to DSL or cable internet end devices used for access.  HGW (Home Gateway): Manages home services such as WiFi, LAN, IPTV, and VoIP. device.  STB (Set-Top Box): A device used for IPTV and digital broadcasts. 25  VoIP ATA / IP Phone: End devices used for fixed telephone service.  WiFi Access Point: Devices that provide wireless network coverage.  Mesh Router / WiFi Extender: Additional home network devices for wider coverage.  Residential Gateway (RG): Integrated router-modem, NAT, and firewall functions. presents. 30  Switch (L2 / L3): Network segmentation and routing components. 18  Power Backup Unit (UPS, inverter): Power for field cabinets or in-home appliances. backup.  IoT Devices (smart meters, home security sensors, etc.): Next-generation fiber / DSL Monitorable devices connected via this method.  Examples of Managed Mobile Network Telecommunication Devices are as follows: Platform 5 It supports all types of mobile devices.  BBU (Baseband Unit): The base that enables the processing of radio signals. station component.  RRU (Remote Radio Unit): A unit positioned near the antenna that processes RF signals. module. 10  DU (Distributed Unit): The processor module at the access layer in 5G infrastructure.  CU (Centralized Unit): Located near the 5G core network, it provides data and control capabilities. The governing unit.  Small Cell / Femtocell: Mobile cells used to increase local coverage. cells. 15  eNodeB / gNodeB: Main components of LTE (4G) / 5G base stations.  Mobile Core Network (EPC / 5GC components): MME, SGW, PGW, SMF, Core network administrators like UPF.  IoT Gateway / NB-IoT devices: Mobile network over narrowband. Connected devices. 20  Mobile Hotspot devices: End devices that provide 4G / 5G connection sharing.  SIM card modules (eSIM platforms): Enables remote provisioning. integrated SIM systems. Each component agent architecture included in fixed and mobile network devices (406) They are tools. The platform that is the subject of the invention has 25 features in addition to those written in my application document. There are numerous different fixed and mobile network devices (406). Other Artificial Intelligence Application Tools (407): As required by the platform These are external AI-based services or tools that can be called upon. For example, a machine translation service, image processing model, multi-purpose generative artificial intelligence model or Tools such as optimization algorithms are included in this scope. Predefined task agents 30 (304), such AI tools can also be used via vehicle caller (305) when needed. can use. 19 Each component agent included within the scope of other artificial intelligence application tools (407) They are tools in architecture. Databases (408): Internal databases (408) managed by the telecom operator (804) and for example, data outside the company such as government records, third-party customer (803) data are the databases (408). The system accesses these databases (408) via connectors (404) 5 by accessing, it retrieves or writes information as needed. Each within the scope of databases (408) In a component-agent architecture, they are tools. External cloud services (409): Open and private services required by the telecommunications operator (804). Digital storage systems and computing infrastructure procured from cloud service providers, These are cloud services such as databases, web servers, and load balancers. External cloud services 10 Each component included in (409) is a tool in the agent architecture. Servers (410): Located in the infrastructure where the telecommunications operator (804) provides service. Server equipment. Servers (410) hardware, hardware-specific operating system, purpose It consists of appropriate software. The web where the software for each module of the platform runs, As application servers, the large language model (301) and the 15 agents run on it. There may also be numerous GPU-powered computing servers. Servers (410) Each component found is a tool in the agent architecture. Intelligent operation module (501): Detects anomalies, errors, and performance issues in the system. artificial intelligence models that identify and suggest solutions to problems, and these models It is a module consisting of analyses. It is an external component integrated into the platform. This module has 20 in all hardware and software components, subscriber / customer (803) / user transactions, By detecting anomalies in events occurring both inside and outside, the platform reports the incident. It provides information via the trigger. This module is descriptive (describes the past), diagnostic. (explanatory), predictive (forecasting the future), and preventive (suggesting solutions) It works with numerous machine learning and deep learning approaches. This module covers 25 machine learning and deep learning applications. It uses models. In the input that the intelligent operation module (501) will provide to the platform Event trigger module (103) planning agents according to a change that will occur (302) activates. The intelligent operation module (501) activates the event trigger module (103). It provides detailed information about its findings in a specific format. Structured data (502): Structured data such as database tables, charts, and records. 30 For example, data such as subscriber lists, traffic statistics, or device inventory are structured. Examples of data sources include the large language model (301) and agents that use data from these sources. By inquiring, they learn about the current situation. Unstructured data (503): Documents, emails, log records, technical documents, Free text or documents with an undefined structure, such as team texts. These are data in the following formats. For example, a device's log files, learned lesson reports, or 5 Texts such as failure history are included in this scope. The large language model (301), RAG (Retrieval- By extracting relevant information from such documents with the help of Augmented Generation architecture. can analyze. Or, using unstructured data (503), the large language model (301) can be examined. It can be adjusted. Knowledge base (504): Contains structured information and enables agents to make informed decisions. It is a data repository that provides rules, facts, relationships, and history. It may consist of experiences. Agents use this knowledge base (504) while performing their duties. They use them in questioning, drawing conclusions, or planning processes. Especially in subordinating tasks. cognitive skills such as task division, reasoning, decision-making, planning, and evaluation. In tasks, the knowledge base (504) is critically important. Thus, agents respond to environmental inputs 15 Instead of simply reacting, they can make informed, rational choices. Information found in the knowledge base (504) may be, for example, as follows: The subject of the invention The system may contain much more diverse information.  Network and Device Information Technical specifications of network devices (OLT, ONT, RRU, BBU, Switch, Router, etc.) 20 Device compatibility tables and configuration templates (modem, set-top box, (HGW etc.) Frequency plans, cell structures, geographic coverage maps those IP address blocks, VLAN definitions, QoS profiles Topology and connection rules for 5G / 4G / FTTH architectures 25  Service Definitions and Business Rules Subscription packages and campaign information (data limit, speed, pricing) Service activation / deactivation rules Billing and collection rules (prepaid / postpaid scenarios) Service levels (SLAs) and quality criteria (QoS / QoE) 30  Troubleshooting and Support Information Common failure scenarios and troubleshooting guidelines Troubleshooting guides for customer support teams 21 Solutions suggested according to complaint types. Automatic fault diagnosis rules (fault trees, root cause analyses)  Customer Profile and Behavioral Data Customer segmentation and behavioral models Service usage analytics (e.g., which regions have high data usage) 5 Preferred channels (mobile app, call center, dealer, etc.) Frequently Asked Questions (FAQ) and Answers Ontology and semantic concept maps (e.g., “changing packages = tariffs”). update”) 10 Labeled dialogue examples for NLP models Rule sets to be used in planning and decision-making processes.  Regulations and Policies that is, the Personal Data Protection Law (KVKK), BTK regulations, and subscription contract rules. Roaming agreements Security, access and authorization policies 15 Synthetic data (505): Used in system development and training processes, derived from real data. These are artificial data created with derived or fictional scenarios. The large language model... (301) It is used for training or testing agents. In this way, the model is used for rare occurrences. or becomes prepared for critical scenarios. Telecommunication memory (506): A memory and information store within the system. It is a component. On the one hand, the previous question-answer contexts of the large language model (301), It allows the system to remember user session history and actions performed (such as dialog history). On the other hand, by also storing specialized knowledge and experience in the telecommunications field. It provides contextual information to the planning or large language model (301) components when needed. This memory also serves to share information between agents. It exists in both the long and short term. available. Business process pool module (601): Tasks of planning agents (302) in a business process; when dividing into subtasks, prioritizing and ordering subtasks, and using predefined methods. In order to make the most accurate decisions when matching with task agents (304), reference acceptance These are definitions containing the administrator's (801) advice, suggestions and directives regarding the work process. 30 A business process pool provides detailed definitions for the successful execution of a business process. This is done for each business process in the module (601). Each business process pool module (601) Steps to implement a business process detailed by admin (801) 22 This enables planning agents (302) to identify situations they encounter for the first time. or being able to plan and reason better in situations where improvement is needed, and In order to make a decision, the previous Evaluation and Feedback Agents also need to have the information. by reviewing their actions, learning better from mistakes, and changing their strategy For the business process pool module (601), defining the business processes and these processes 5 It needs to be constantly updated. In this way, the definitions in a business process are completed in advance and controlled by humans. Input triggering-work process-task-subtask-task agent-tool matching and work through its capabilities. The process flow is introduced to the system at least once. This allows for continuous improvement of the system. For its operation, business processes can be continuously revised by the admin (801) and 10 It will be updatable. Here are some examples of business processes:  Subscription and Customer Management Processes  Service Activation and Provisioning Processes  Network Management and Operations Processes 15  Customer Experience and Call Center Processes  Inventory and Logistics Processes  Sales and Marketing Processes  Billing and Collection Processes  Legal Compliance and Security Processes 20  Internal Operational Processes  Quality Assurance Processes  Team Performance Monitoring Processes  Team Operations Processes  KPI Tracking Processes 25  Troubleshooting and Fault Prevention Processes  WiFi Optimization and Improvement Processes  Mobile Network Slicing Processes Data lake and repository (701): All transaction logs, collected metrics, and raw data. It is a big data environment where data is stored centrally. Every 30% of structured and unstructured data... Various types of data are stored here for long periods. The data lake allows for advanced analysis of historical data. 23 creating a foundation for training machine learning models and conducting retrospective inquiries. prepares. Monitoring & Logging (702): Every step of the system operation and significant network events It is the component that is monitored and recorded in real time. All user requests, The transactions performed, decisions made, and results are logged and timestamped. It is archived. The monitoring component provides system administrators with the ability to monitor via a real-time interface. While the materials are presented, the records are used for auditing, debugging, and continuous improvement purposes. Counter module (703): For each operation performed on the platform;  how many tokens were used and / or  how long it was used and / or 10  how much resources (processor power, memory, etc.) are consumed and / or  how many service calls were made This is the module that performs the calculations. The calculation method and metrics are different for each agent service. This module holds aggregate or real-time streams stored in the data lake and repository (701) module. It performs calculations using the data. 15 Pricing module (704): Service of each internal and external customer (803) on the platform charging for consumption based on the calculation made by the counter module (703) and This is the module that invoices the customer (803). Admin (801): This role includes those authorized to use the platform. Admin (801) It is the person who notifies or defines the request or action plan to the system. Admin (801) 20 The request is understood through the intent detection module (102) and the planning agents (302) are triggered. Admin (801) business processes flow module (104) business processes will follow one another It can create a chain workflow and trigger planning agents (302) accordingly. Admin (801) Timer module for business processes to be triggered at a specific time (105) You can schedule time and trigger scheduling agents (302). Admin (801) 25 By configuring the web search agent module (106), this module can trigger the system. to conduct appropriate research in the outside world and share feedback Makes the necessary adjustments. Admin (801), platform management module (101) You can update all system components using the Admin (801) job process pool. You can continuously update the steps of the business processes in module (601). 30 24 Developer (802): This role includes those who develop the platform. All agents are pre-selected. It is developed by the developer (802) and by the admin (801) of the Telecom Operator (804). Agents are provided ready for use. Customer (803): End users served by the telecommunications operator (804) or They are subscribers. It is the person who notifies the telecommunications operator (804) of their request. The customer's (803) request, 5 Planning agents (302) are triggered by agreement with the intent detection module (102). Invention The system within the scope of which customers (803) can log in with their own accounts when necessary by accessing limited information or performing some self-service operations He can allow it. Operator (804): performs basic operations on the system, system 10 not performing configuration, but fulfilling responsibilities such as error detection and troubleshooting It is a role. Working principle of the invention A1. Development Step for the Platform's Core Components: The platform is a telecommunications operator (804) or service provider or any sector 15 Steps A.1.1., A.1.2., A.1.3., and A.1.4. must be completed before the system is made available for use by a business. It is applied. The methods given below ensure that all technology stacks are included in the platform during the application process. These are the choices made during development, and different options are also available depending on the needs of the invention. It can be used. In addition, the invention allows the technology to be updated over time. 20 It supports [the system]. It is used according to the changing and developing possibilities in the future. Technologies can be updated. A1.1. The developer (802) develops the following core platform components. From deployment Then there will be no need to develop these components further.  Platform management module (101) 25 It is developed to manage all other components of the platform. This module is a multi-layered web application with a user interface and backend services. It is developed in this way. Designed with a microservice architecture and implemented with other components. It communicates via a programming interface. It can operate in both cloud and local environments. The services are deployed in a container-based manner; the administrative interface is a web-based client. It is presented as a layer. Backend services for high availability are multiple It is deployed with an instance and scaled with a load balancer. For the backend, Java Spring Boot or An enterprise-level, proven framework like Node.js / Express is used. Front-end Modern web technologies such as Angular and React are used for the admin panel. Database 5 a relational data client like PostgreSQL to hold configuration and system metadata The underlying technology is suitable. Docker for containerization, Kubernetes for orchestration. This facilitates the deployment of the module in both the cloud and on-premises environments.  Intent detection module (102) Analyze free text requests from humans or machines using natural language processing. 10 It is developed to determine the user's intent and the details of their request. SpaCy, based on Python, is a natural language processing library like Hugging Face Transformers. or directly existing large language model (301) application programming interfaces It is used. Optionally, for intent classifications requiring deep learning, PyTorch is available. TensorFlow frameworks are preferred. For service delivery, Python 15 such as FastAPI or Flask are used. Web frameworks are used. Alternatively, if the Java environment is preferred; Libraries like DeepLearning4J or Stanford NLP, and REST with Spring Boot. Platform differences are also achieved by adopting an application programming interface approach. It is supported by technologies such as Kafka to handle the request traffic in both cases. Usage is scaled in conjunction with a messaging system. 20  Event trigger module (103) Event trigger module (103) triggers events occurring inside or outside the platform system. any changes are detected, the necessary platform components are alerted, and the necessary actions are taken. It is developed to facilitate the taking of actions. This module is designed with an event-driven architecture. The intelligent operation module 25 on the platform. (501) and monitoring & recording (702) listening to event messages from other components It is developed as a message queue consumer. For example, Apache Kafka or RabbitMQ. It captures events published on specific topics within its infrastructure by subscribing to them. Microservice It is a lightweight and responsive service; multiple instances for high availability. Fault tolerance is achieved by consuming data using this method. Event handling logic, when necessary; rule 30 with an engine or functional programming approach, the rules can be easily changed. It is implemented in this way. This module is a component that specifically requires real-time performance. 26 Therefore, latency is kept low and pressure feedback is used for scalability. A management mechanism is implemented. Real-time systems like Kafka Streams or Apache Flink are used. Stream processing technologies are used for this module. For example, events are processed into Kafka threads. When it crashes, the trigger module runs as a Kafka Stream application and executes specific patterns. It can detect it. For programming languages, you can use Java or Flink 5 for the Kafka Streams library. Scala is used. Alternatively, a lighter framework like Redis Pub / Sub with Node.js can also be used. It can be implemented, but in the telecommunications scenario, a scalable technology like Kafka is preferred. It has been developed. A rule engine like Drools is integrated to define event rules. Deployed as a Docker image in a container environment and automatically runs on Kubernetes. Restart and scaling support is provided. 10  Business process flow module (104) This module enables multiple business processes defined by the admin (801) to trigger each other. It incorporates chain flows. The start and end of each business process are recorded on the platform. It acts as a trigger. This component is designed as a workflow management system or rule-based engine. This 15 The module is developed specifically for the platform as a microservice-based orchestrator; Camunda, Flowable, etc., are examples of BPMN (Business Process Model and Notation). (and Representation) based process engine is also an alternative, depending on the business's preference. It is used. In accordance with the event-driven structure, it initiates a business process or the next one. The next step is triggered by `-20`. This module stores workflow definitions and `-20` It includes state machine logic for execution. Stateful for high availability. It is stored in a shared repository, such as a database, and multiple instances are run in parallel. Failover is enabled. Services are designed to work both in the cloud and on-premises. by containerizing the necessary resources so that they can be scaled dynamically. is designed. 25 For code-based solutions, consider Java, Spring Boot integrated with Spring State Machine, or Netflix. Orchestration libraries such as Conductor are used. Alternatively, Camunda, Zeebe, Open-source BPM systems like Activiti are used. These systems use workflow models (BPMN). It can be defined and run as version 2.0. PostgreSQL for secure and consistent data management. Flow states can be stored on it. Kafka integration for messaging, workflow 30 It enables asynchronous communication between the steps. All these components are Docker containers. packaged and managed on Kubernetes, ensuring consistency across different environments. Work is provided. 27  Timer module (105) Timer module (105) schedules specific business processes at predefined times. It is designed to start automatically. This module creates time-based triggers. It is designed as a service. Traditionally similar to cron; Java-based Quartz. The library can use a scheduler like cron, or, according to customer (803) preference, cloud 5 To be compatible, serverless scheduling applications such as AWS EventBridge and Azure Logic Apps are used. Services are also integrated. In local deployment, a timer is embedded within the application. Alternatively, Kubernetes CronJob can be used. For high availability, a single schedule A leader selection method where one source is active and other examples are in reserve. It is used as an independent "scheduler" service within the microservice architecture. It works and sends a trigger signal to the planning agents (302) via message. Event-driven design As required, it initiates the relevant processes by publishing an event at the scheduled time. Quartz timer solution for a Java-based application integrated with Spring Boot. It is used in this way. In the Python ecosystem, libraries similar to APScheduler are used. Because containers are used throughout the platform, calendar triggers are in container 15. It is configured to operate in a time-sensitive manner. Scheduling in the database. PostgreSQL is used for recording and modifying tasks as needed. For relatively lighter tasks, SQLite is used. Alternatively, CloudWatch can be used in the cloud. Services like Events or Azure Scheduler can also be used as triggers.  Web research agent module (106) 20 The WEB research agent module monitors developments in the outside world and focuses on a specific topic. so that the platform can take the necessary actions when a change or new information emerges. It is the component that triggers the system. This module is designed as an agent-based background service. This module also... It is also used as a tool to perform a specific operation over the web. Continuously or 25 It is planned to periodically scan external resources. In a microservice architecture. It is a standalone service and can acquire tasks by pulling tasks or through a streaming subscription. For example... RSS feeds can be accessed via an application programming interface or by specifying particular feeds for web pages. It can collect data by sending periodic requests. This module is used for web scraping and application programming. Since resource-intensive operations such as interface calls are performed, parallel processing is required when necessary. It operates efficiently using particles or asynchronous requests. When using the cloud, specific also as a serverless function, such as AWS Lambda tasks that are triggered at intervals 28 It is considered. In situations requiring continuous monitoring, a permanent service is continuously available. High For accessibility, multiple agent instances are deployed to monitor different resources; For scalability, each monitoring target is replicated horizontally as a separate business unit. Web scraping can be done using libraries like Scrapy, BeautifulSoup4, and Selenium, all within the Python language. and is used in data retrieval. SERP application 5 for retrieving search engine results. SearXNG is used as an open-source alternative to the programming interface. External Requests / HTTPX or gRPC clients are used to retrieve data from APIs. Node.js It is a suitable alternative with its asynchronous structure; for browser automation on the Node side. Tools like Puppeteer and Axios are used for HTTP requests. The collected data is stored. or a NoSQL database like MongoDB to detect change and 10 for status tracking Redis is used. Machine learning is applied to the keywords or patterns being tracked. If necessary, integrate a scikit-learn or TensorFlow-based solution into a Python environment. It operates as such. This module is also distributed within a container, ensuring platform-wide compatibility.  Platform channel module (107) The platform uses people, software, and hardware to exchange information. Any improvements that will enable one-way communication are made in this module. Platform channel module (107) is specialized for managing different communication channels. It consists of many components. Each component is responsible for a specific type of communication, incoming and outgoing. It connects the outgoing data to a standard interface. This allows the module to communicate with the platform in various ways. It acts as a bridge between environments. Architecture, with an agent-based approach, 20 It is designed so that each sub-component within the module has its own area of ​​expertise (e.g., audio). It acts like an agent focused on (search, chat, IoT data, etc.) and is the core of the platform. It connects to the system using defined protocols. Thanks to this distributed structure, communication channels Flexible coupling is provided between them, and any single channel component is independent of the others. It can be updated and scaled accordingly. The platform channel module (107) essentially supports different systems 25 It functions like a communications hub, bringing together all of the platform's features under one roof. It allows for the consolidation of channels. The architectural approach treats each communication channel as a separate service or microservice. For example, a component dedicated to voice calls (a telecommunications interface or VoIP). While there is a gateway, there is a different component for text chat and messages (chat server 30 (or message queue consumer) is found. Similarly, application programming REST / GraphQL services for managing interface requests from IoT devices. 29 Specialized units like MQTT / CoAP brokers for processing data are part of the architecture. These are the components. All these subsystems are coordinated under the Platform channel module (107). By working with the platform's core functions, it enables event-based or request-response-based communication. They transmit data via this method. Thus, the architecture ensures reliable data transmission even at high concurrency rates. It provides near real-time streaming. This distributed and modular design enables the platform's communication 5 It provides scalability and flexibility at the layer, while also ensuring that each channel It allows for the use of the most suitable technology according to their needs. The platform channel module (107) includes a wide variety of communication technologies and protocols. It is built with a rich technology stack that encompasses different types of platforms. It is designed to enable interaction across channels. The main technology is 10. Its components are as follows: Voice Communication: The platform supports interaction with classic voice networks and VoIP technologies. For example, the module can be integrated into a VoIP / SIP infrastructure or a cloud-based call center platform. It is integrated. Thanks to WebRTC support, real-time communication is possible via web and mobile applications. Time-based voice / video calls can be made. This allows the platform to handle phone calls up to 15 seconds. It meets real-time voice communication needs such as conference calls. Call center applications include human-based call center and AI-based call center. It is divided into two categories as a central application. In artificial intelligence-based call center applications... VoIP / SIP infrastructure, text-to-speech and speech-to-text artificial intelligence. It is integrated into the models. At the same time, there is a large language model (301) 20 specialized for this purpose. Artificial intelligence handles the planning and reasoning tasks of the call center. Artificial intelligence The call center, based on this platform, is also an agent built with its capabilities. Written / Chat Communication: Platform channel module (107), instant messaging and chat It also supports various channels. Users can access them via mobile or web interfaces. You can have instant chats with the platform; this communication takes place in the background via a continuous bidirectional system like WebSocket. This is provided via connection protocols or messaging protocols such as XMPP. Email and Traditional text-based channels like SMS are also part of this technology stack; the platform is a It manages incoming / outgoing messages through these channels via an SMTP server and SMS gateway. Web Services and APIs: Various APIs for integrating external software with the platform. Communication styles are supported. These include REST (HTTP-based traditional web 30 services), GraphQL (which enables the client to retrieve exactly the data it needs) query language), gRPC (high-performance binary RPC protocol), SOAP (XML-based traditional integration), WebSocket (continuous server-client connection), and There are technologies such as WebRTC (real-time media communication). Also... By supporting modern web communication protocols such as HTTP / 2 and QUIC, it provides more... Fast and secure data transfer is enabled. Thanks to this rich API layer, the platform, It also communicates seamlessly with microservice architectures and enterprise applications. 5 IoT and Device Communication: For communication with devices in the Internet of Things (IoT) ecosystem. The module uses lightweight and efficient protocols. Specifically, MQTT (Message Queuing Telemetry). IoT protocols such as CoAP (Constrained Application Protocol) and CoAP (Transport). By providing support, data from low-power devices or closed networks can be brought to the platform. This enables the transmission of data. These protocols are suitable for sensors or devices with limited bandwidth and resources. It enables reliable and efficient data exchange between devices. Platform channel module (107), collecting data in IoT integrations with the publish / subscribe model and It enables commands to be transmitted to devices. Message Queues and Real-Time Data Streaming: The platform enables high-scale data flow and It also includes messaging queue technologies for event processing. Specifically Apache Kafka and 15 Platform components are integrated with popular messaging systems such as RabbitMQ. Asynchronous data transfer and event propagation are performed between them. Kafka is a high-volume system. While used for real-time data streams, RabbitMQ provides reliable messaging with low latency. It is preferred for transmission and complex routing scenarios. This allows the platform channel module (107) can process millions of events in seconds, both high data transfer and 20 It has an infrastructure that supports scenarios requiring low latency. This message Focused architecture enables loosely contextual communication between services within the platform and the system It increases scalability.  Adapter (401) The Adapter (401) module connects the agents on the platform with external business support systems (402) and 25 These are interface components that enable communication between operational support systems (403). Its primary function is to convert standard API calls into the protocols and formats of these systems. It is about transforming. Since adapters (401) contain a structure specific to the system they are integrated into, each target system or configured as a separate microservice for the protocol. For example, a customer relationship 30 A management system adapter can be divided into different services, such as a network management system adapter. Some adapters are created as add-ons. Adapters (401) are generally synchronous on demand. 31 It works; when it receives a request from the caller (305), it immediately transmits this request to the target system. It returns the answer. If the target system call is a long operation, adapter(401) Asynchronous mechanisms are implemented within it, such as leaving a message in a queue and voting on the response. Since adapters perform (401) protocol conversion, security and data integrity are critical. These are the issues. For example, an adapter (401) that will execute a CLI command, 5 within the command. It must process the parameters securely and execute the command. For high availability, each There are multiple instances of the adapter (401) service and a specific adapter (401) If it malfunctions, pending requests are redirected to another one. The adapter (401) is used as a module in Apache Camel. The components support many protocols such as SNMP, SOAP, JDBC, FTP, and Telnet. Every 10 The adapter (401) service defines Camel redirects and directs the incoming REST call to the relevant It maps to the protocol. According to the admin's (801) preference, for solutions other than Java, Node.js And adapters (401) can also be written in Python. For example, SSH connection with Paramiko in Python. An adapter (401) is developed that installs and sends CLI commands. To give some protocol examples. For example: SNMP4J on the Java side, easysnmp on the Python side, and 15 on the Java side. For SOAP, use JAX-WS, for Python use zeep, and for CLI use use Apache CommonsNet or... The shell exec is used directly. For example, the API key of the OSS system, the BSS database... Adapters (401) require user information, such as system authentication. It is also developed to be compatible with their methods. This information is, for example, used with a solution like HashiCorp Vault. It is stored securely and this information is retrieved when accessing the adapter (401). 20  Business process pool module (601) This module allows planning agents (302) to run a business process for the first time or by the manager so that they can make the most accurate decisions in situations that need improvement. to provide the provided guidelines to the platform and the format defined as business processes It is developed to guide planning agents (302). 25 A business process repository is a type of repository or catalog. Business processes are stored in a database table. or stored in configuration files. It contains the name of the process, its purpose, steps, and each... Which of the predefined task agents (304) are recommended to be used in step 3? JSON / YAML formats or text files that hold information about what will happen in the system; admin (801), using the platform management module (101) interface, each business process 30 It is developed for the purpose of reconfiguring and continuously updating the system. 32 In a PostgreSQL database, a table like "is_surec_tanimlari" is used, with a JSON column. Process details are stored. The same information is also stored in a NoSQL database such as MongoDB. It can be held.  Counter module (703) The counter module (703) is event-driven and complies with microservice architecture principles. It is configured. Each transaction on the platform can be processed at the time of the transaction or in batches. These records are received by the counter module (703) and fed into the relevant calculation algorithms. is directed. The module is configured to include the following subcomponents:  Data Collector: Collects event logs from different points of the platform and 10 It is the component that performs preprocessing. A Kafka or similar message queue. It retrieves data from its infrastructure.  Tracker and Calculator Engine: Specific agents are assigned based on the collected events. It is a customizable processing layer that calculates metrics.  Timer and Batch Processor: In addition to real-time calculations, 15 It is a component that performs batch calculations at scheduled intervals.  Data Router: Data lake and repository (701) module of calculated metrics It enables the writing process. If necessary, it also generates trigger signals for warning systems.  Architectural design, high scalability, independent component updatability, and It is created with data processing priorities with minimum latency. 20 The Go language is suitable for high-performance metric computing components, flexible data processing, and batch processing. For analysis tasks, the Python language is integrated with data processing engines such as Spark or Flink. Scala or Java languages ​​are used in its implementation. In the messaging infrastructure, event-based data is used. Apache Kafka is used for streaming and process triggering. As a data processing framework, it is actually... Apache Spark Streaming is preferred for real-time and aggregate metric calculations. Instantaneous 25 Prometheus as a supporting tool for metric monitoring and time series analysis. Hadoop HDFS is used for data storage, while InfluxDB is used for time series. Databases such as TimescaleDB (408) are used. gRPC is used for API and service integration. The REST API allows for low-latency and structured data exchange with other modules. It is used. 30  Pricing module (704) 33 The billing module (704) charges each internal or external user of services on the platform. By monitoring the customer's (803) transaction-based resource consumption, the corresponding cost It is the module that performs the calculation. The basic function of this module is the counter module (703) consumption data measured by (number of tokens, hardware resource consumption, software Based on factors such as license, number of users, usage duration, etc., 5 predefined service fees of customer (803) in accordance with pricing policies The goal is to calculate and trigger the billing processes. This module includes service-based pricing, user segmentation, and usage time-based pricing. multi-faceted, including fee calculation, resource-based pricing, and quota overage charges. It is designed with the flexibility to support pricing rules. Furthermore, this module is commercial 10. Reportable and traceable invoice items in terms of transparency and customer (803) audit It also bears the responsibility of production. The billing module (704) is microservice-based, event-driven, and rule engine-supported. It is developed with a structure. This structure modularly accommodates different pricing scenarios. managing subscription-based or usage-based services independently of each other 15 It makes it possible to distribute the evaluation and billing logic in a decentralized manner. He / She / It does. The architectural components are as follows:  Consumption Data Receiver: transaction issued by counter module (703) It is the component that retrieves metrics by listening to message queues. 20  Fee Calculation Engine: Calculates fees for each transaction or service according to pricing rules. This is the core algorithm layer that calculates the cost for the session.  Rule Definition System: the admin (801) pricing rules, service management that defines exceptions at the level and time-based campaigns It is a panel. 25  Billing Interface: Calculated charges into customer (803) based invoice items It is the layer that converts and integrates with external billing systems. A1.2. Developer (802) develops the following artificial intelligence components of the platform:  Large language model (301) 34 Detecting intent, understanding natural language input in the environment, and producing appropriate outputs. general usage such as producing, analyzing and generating content like code, reasoning for their purposes, Moreover; o Planning agents (302); Analyzing tasks, breaking them down into subtasks, sub 5 Assign predefined task agents (304) to tasks Evaluating the task quality of the Assessment and Feedback Agents, analyze errors, generate feedback o Predefined task agents (304) produce specific outputs, with tools The study focuses on generating context-appropriate responses. 10 In order for it to fulfill such responsibilities, the platform has been created for that purpose. One or more major language models (301) are developed and used. The major language model (301) used on the platform that is the subject of the invention was not changed as well as for specialized tasks;  full fine tuning, 15 • Fine-tuning of the part,  Parameter-enabled fine-tuning Low-rank adoption. Quantized low-order tuning, that Adapter (401), 20 Prefix and request adjustment, BitFit It can be restructured using methods such as these. For model development / fine-tuning, use the Python descriptive language PyTorch or TensorFlow. It was used. To service the model, a FastAPI or gRPC-based server or 25 TensorFlow Serving or Triton Server frameworks are preferred. Hugging Face models. Model optimization is done using the Transformers and Accelerate libraries when applicable. This is done. FAISS, Milvus, and dense vector ElasticSearch are used as vector databases. This module is computationally intensive, therefore it requires CUDA and related GPU drivers and libraries. It is also used. The container image is built on NVIDIA's NIM-based services. Big 30 language model (301), any model as open source and closed source The platform includes multiple-choice options. Each agent's role is based on their activity and the task they will perform. 35 which language model to use depending on its size, success, and resource consumption It is determined during the design and development phase. The Big Language Model (301) requires a separate AI service due to its high computational needs. It is considered as a set. To fit the microservice architecture, the model itself is a It is hosted on the service and accessed by the other 5 via the REST / gRPC application programming interface. It provides services to the components. This service also runs on servers with GPU support, and It scales. For high availability, a copy of the model is available for active-active inference services. It can be configured. Due to the large model size, load balancing and request queuing are necessary. Mechanisms are used. When cloud computing is used, such as SageMaker and Azure ML Endpoint are employed. Services are used. For local servers, Nvidia Triton Inference Server or 10 HuggingFace Serving and similar solutions enable distribution within containers. Also, large language options are available. around model (301), for long-term memory and knowledge base (504) integration RAG (Retrieval-Augmented Generation) architecture that will work together with a vector database. is created. For the specialized task of each module and agent in the invention, numerous and diverse major languages ​​are used. Model (301) can be used. Depending on the intended use in the invention, it can be called a Small Language Model or a Visual Language Model. Model (Vision Language Model), Expert Selection Model (Mixture of Expert), Big Mind Language models based on different concepts, such as the Large Reasoning Model, are also major It can be used instead of the language model (301). In the invention, the large language model (301) is represented as 20 It is used here in place of everything written or not written. The invention ranges from which language model is used to how the language model is fine-tuned from the outset. from the implementation or partial fine-tuning, which language, framework and language model the language model uses It is independent as it was developed with the library. Its purpose is based on the technological conditions of the day. The language models and development methods used vary accordingly. 25  Planning agents (302) These agents facilitate task coordination and workflow in multi-agent AI architectures. autonomous systems used for guidance and the identification of appropriate sub-tasks. It is developed with the aim of being a decision support component. Planning agents (302), from triggers Analyze incoming information and business process directions (601) in the business process pool module 30 By doing so, it identifies tasks and sub-tasks that will ensure the fulfillment of business processes. 36 It configures the system by distributing the appropriate predefined task agents (304). It enables it to work in an intelligent and dynamic way. Planning agents (302) only sub not only the order of tasks, but also the priority between these sub-tasks, by also considering dependency and timing relationships, a flexible and adaptable approach. It is developed to generate flow. 5 This agent component is a microservice within the framework of modern software development approaches. It is developed with architectural principles. Planning agents (302) are scalable on their own, as a service that can be independently versioned and operate separately from other agents It is constructed. With a Domain-Oriented Design approach; subscription management, troubleshooting, network optimization 10 There may be separate planning agent services for each major business area. These agents are event-driven. They receive and process job descriptions from a message queue because they are triggered as such. can begin. Each planning agent service includes a rule engine or decision-making engine. It contains an algorithm. Examples of multiple planning agents (302) for high availability. Consumers can work together as a group to share the incoming task queue. Additionally, different 15 For different task types, different queues or threads can be used to divide the workload. It analyzes the trigger commands it receives as input in an event-based manner and generates its outputs again through events. It structures itself with an event-driven architecture that it can broadcast through. In this context, Planning agents (302) also act as orchestrator agents, performing tasks It can also function as a central coordination layer where everything is controlled holistically. 20 It does not store system states directly within itself, but instead uses a central memory. The system retrieves and writes context through a module (e.g., Redis or vector memory). It contributes to its distributed and resilient nature. Technologically, the Python language was used in the development of the planning agent (302), natural language processing 25 and rich library support for large language model (301) integrations It was preferred because of its fast and asynchronous communication with the outside world. The FastAPI framework, which provides application programming interface management, is used. Planning messaging between agents (302) and other agents or system components and Distributed message queue solutions like Apache Kafka can be used for synchronization. 30 LangChain for integration with the Big Language Model (301), prompt generation, memory management and tasks such as vehicle guidance via predefined task agents 37 It is available. In addition, it allows for the permanent management and traceability of task planning outputs. For this, a relational database (such as PostgreSQL) may be preferred. Vector search Vector databases such as Weaviate, FAISS or Qdrant for situations requiring (408) It is integrated into the system. Planning agents (302) other predefined task agents (304) and 5 within the system By communicating asynchronously with the evaluation agents and feedback module (303), one of the fundamental building blocks of a flexible, reliable, and scalable artificial intelligence architecture It creates.  Evaluation agents and feedback module (303) These agents, within multi-agent artificial intelligence systems, execute planned tasks and generate 10... Monitor the quality of the output, generate meaningful feedback, and adjust the process when necessary. It is developed as an integrated intelligence component structured for the purpose of guidance. These agents, It lies at the heart of the system's autonomous decision-making and self-healing capabilities, and Analyzing data obtained during critical stages such as planning, implementation, and output production. It undertakes quality control, fault detection, and corrective guidance tasks. 15 The primary development purpose of these agents is to both predefined mission agents (304) and the accuracy, consistency and target of the outputs of the planning agents (302) By verifying its suitability, the system demonstrates agile and reliable artificial intelligence behavior. The aim is to enable the display of evaluation agents and feedback module (303) It not only determines whether the tasks have been successfully completed; it also 20 By conducting critical analyses of the outputs, it offers suggestions for improvement and, if necessary, Provides feedback to the planning agent (302) for the restructuring of tasks. In this respect, it represents the cognitive control and introspection functions of the system. From a software architecture perspective, evaluation and feedback. These agents, like other agents, are developed using microservice architecture principles. This 25 Thanks to its structure, it can be positioned independently, scaled as needed, and They can be integrated in a flexible structure connected to other components of the system. Feedback Since the process mostly operates asynchronously, an event-driven architecture is preferred. When the task is complete, the relevant agent generates a "task completed" event; evaluation The agents and feedback module (303) listen to this event and check the output to create a 30 The system publishes a "feedback ready" event. This allows for flexible, interconnected system workflows. It is continued. 38 The technology stack proposed during the development process is also selected in accordance with this architecture. Python as a programming language, language model interactions and natural language output analysis. It is the preferred choice because it offers rich libraries for application programming. For interface and integration needs, lightweight, asynchronous, and high-performance options like FastAPI are available. frameworks are used. This agent calls the Large Language Models via the request, 5 It generates feedback and improvement comments. For this purpose, LangChain or a similar platform is used. through frameworks, via request templates structured with Large Language Models Analysis can be performed. In order to evaluate task outcomes in a vectorial context, past data can be considered. To retrieve contexts, you can use a vector database like Weaviate or Redis-supported database. A central memory solution is integrated into the system. Kafka or 10 is used for event-based messaging. A communication infrastructure like RabbitMQ is used to manage the flow of notifications between agents. It is used.  Predefined mission agents (304) These agents are designed to perform specific tasks within multi-agent artificial intelligence systems. Structured, operationally focused, and actionable artificial intelligence 15 These are the components. Predefined task agents (304); planning agents (302) They work to carry out the sub-tasks identified during the planning phase. Their tasks It performs its functions directly, while integrating with external services when necessary. autonomously performs operations such as execution, data acquisition, querying, production, analysis, and transformation. They perform. MCP clients (MCP 20) on predefined task agents (304). It hosts clients. This allows these agents to perform sampling and inference. It performs the building block functions of elicitation and logging. The primary goal of developing these agents is to create high-performance, task-optimized solutions for specific mission types. The goal is to create efficient, action-oriented modules with predefined task descriptions. These agents exhibit flexible artificial intelligence behaviors, as well as deterministic and 25-degree system behaviors. It also makes it possible to produce repeatable outputs. For example, interference at certain levels. The relevant tool allows the agent to remotely make changes to the existing WiFi channels. These agents are used for tasks such as running a function and writing records to a database. It is specifically defined and executed. This structure increases the modularity of the system while facilitating tasks. It enables distributed, parallel, and synchronous execution. 30 From a software architecture perspective, predefined task agents (304) also in accordance with microservice architecture principles in a way that is compatible with the rest of the system components. Each task agent is designed as an independent entity with its own functional responsibility. 39 It can be deployed as a service. This structure enables task-based scalability. For example, a specific type of agent under load (e.g., a "data extraction agent") can act independently. It can be replicated. Event-driven architecture is preferred for task triggering and results sharing. Tasks are marked with events such as "task assigned", "task started", "task completed". should be managed. Where necessary, these agents can be used with an enhanced request method. 5 It can produce outputs supported by the large language model (301), but mostly only external tools. It performs the operation using. Predefined task agents (304) are developed in Python. Python is a wide range of artificial intelligence. Thanks to intelligence and integration libraries, these agents can perform various tasks. It makes it easier to implement. A high-performance and asynchronous framework like FastAPI, 10 agents can communicate with external services based on application programming interfaces. It provides. Depending on the agent's task, via the vehicle caller (305), adapter (401) using; business support systems (402), operations support systems (403), external software applications (405), fixed and mobile network devices (406), other artificial intelligence applications numerous 15 tools, databases (408), external cloud services (409), servers (410) It is integrated into private vehicles. Predefined task agents (304), planning agents (302) and evaluation agents and is developed to work integrated with the feedback module (303). Planning receives a mission order from its agents (302), carries out the operation and evaluates the result and It transmits feedback to the agents or the central log system. 20 regarding the outputs of the tasks. Feedback enables the system to evolve into a learning and developing structure. In this context, predefined task agents (304) represent the operational intelligence of the system and specific It forms the core AI workforce that reliably performs its functions. Defined Their ability to perform operations with high accuracy within their operational limits enables them to perform all tasks. This makes architecture indispensable in terms of sustainability and operational efficiency. 25  Vehicle caller (305): Model Context Protocol (MCP) servers (410) cluster It is a custom integration layer that runs on the (MCP Server). Predefined task agents (304) need to use external tools and system functions appropriately. 30 as an intermediary component that enables calling via the application programming interface. All sub-task actions necessary to implement telecommunications business processes are developed. This is triggered in the relevant tools via this module. In short, it activates agents within the platform and external agents. BSS / OSS systems, network devices, databases (408) or external services worldwide 40 It establishes a context between them and performs the operation using the correct protocol. This The module standardizes interfaces of different systems with different protocols and formats. Because it was developed to be able to call upon such things, it serves the purpose of "single-point access to all tools". Tools, resources and requests are available thanks to the tool caller (305) module. (prompts) basic building block characteristics are provided. 5 Vehicle caller (305), task agents programming specific vehicles or applications. executable required to call functions defined on interfaces It provides the executable context. This context depends on the type of function to be called. parameter structures, permission levels, task intent, and the user's position within the system Dynamically generated according to profile features. Predefined task agents (304), 10 to perform the tasks for which they are responsible to the vehicle caller (305) component It accesses; this component also uses appropriate application programming to call the relevant function. interface (API), software development kit (SDK), command-line interface (CLI), or service endpoint It triggers the endpoint via the MCP protocol. This structure is not just a simple proxy server or routing layer; it also includes 15 context standardization, function signature validation, parameters critical operations such as matching and implementing security restrictions It functions as an active context processor, undertaking predefined tasks. their agents (304) do not directly, but in an isolated manner, with all means throughout the system. Interaction is enabled through a secure layer. 20 The vehicle caller (305) is designed with high availability and context processing performance in mind. It is deployed in a microservice-based, horizontally scalable architecture. Each MCP server has specific features. It is associated with a family of tools or a set of functions, and the function definitions, tool capabilities, It includes contextual configurations related to access conditions. These servers (410) function distribution, context solver and tool adapter (401) subcomponents make up a 25 It operates with an orchestration structure. Contexts are dynamically generated based on sub-tasks, and this Calls are executed in accordance with the "stateless" principle. All components form a service network structure. It can operate within, thus providing observability, security, and disabling. Their strategies can be managed centrally. This structure reduces the load on task agents and simplifies the function call logic. By standardizing, it increases the sustainability of the system. Furthermore, planning and evaluation... to perform call routing based on contextual decisions of agents 41 Event-driven design with communication layers like Kafka, NATS, or RabbitMQ. It can be integrated. After the platform is put into operation; Big language model (301), planning agents (302), evaluation agents and feedback module (303), vehicle caller (305) components after commissioning again 5 No further development is required. These components are added to the platform by admin (801). when a new business process, a new task, a new subtask, or a new tool is included will be configured; for example, the large language model (301) will be fine-tuned or RAG New information source components will be added to the (Retrieval-Augmented Generation) architecture. When a new business process, a new task, a new subtask, or a new tool is added to the platform, 10 developer (802) subtask-specific predefined task agents (304) and vehicle caller (305) fundamentally improves the MCP servers held by the party. A1.3. Developer (802) integrates the following platform components with each other according to the design. and completes integration improvements;  Other components integrated with the platform management module (101); 15 o Intent detection module (102) Event trigger module (103) Business Process Flow Module (104) Timer module (105) Web research agent module (106) 20 Platform channel module (107) o Input security layer (201) o API gateway (202) o Identity and access management (203) o Great language model (301) 25 Planning agents (302) Evaluation agents and feedback module (303) Predefined task agents (304) o Vehicle caller (305) that Adapter (401) 30 Business support systems (402) Operation support systems (403) 42 that Connector (404) o External software applications (405) Fixed and mobile network devices (406) Other artificial intelligence application tools (407) those databases (408) 5 External cloud services (409) those Servers (410) Smart operation module (501) o Structured data (502) o Unstructured data (503) 10 o Knowledge base (504) Synthetic data (505) Business process pool module (601) o Data lake and repository (701) o Monitoring & Recording (702) 15 Counter module (703) Pricing module (704)  Other components integrated with the intention detection module (102); Platform management module (101) Event trigger module (103) 20 Business Process Flow Module Web research agent module (106) Platform channel module (107) o Input security layer (201) o API gateway (202) 25 o Identity and access management (203) o Great language model (301) Planning agents (302) o Unstructured data (503) o Knowledge base (504) 30 o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703)  Other components integrated with the event trigger module (103); 35 43 Platform management module (101) o Intent detection module (102) Business Process Flow Module (104) Web research agent module (106) Platform channel module (107) 5 o Input security layer (201) o API gateway (202) o Identity and access management (203) o Great language model (301) Planning agents (302) 10 Business support systems (402) Operation support systems (403) o External software applications (405) Fixed and mobile network devices (406) Other artificial intelligence application tools (407) 15 databases (408) External cloud services (409) Smart operation module (501) o Data lake and repository (701) o Monitoring & Recording (702) 20 Counter module (703) Pricing module (704)  Other components integrated with the business process flow module (104); Platform management module (101) o Intent detection module (102) 25 Event trigger module (103) Timer module (105) Web research agent module (106) Platform channel module (107) o Input security layer (201) 30 o API gateway (202) o Identity and access management (203) o Great language model (301) Planning agents (302) Smart operation module (501) 35 44 Business process pool module (601) o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703)  Other components integrated with the timer module (105); Platform management module (101) Business process flow module (104) o Input security layer (201) o API gateway (202) 10 o Identity and access management (203) o Great language model (301) Planning agents (302) Smart operation module (501) o Data lake and repository (701) 15 o Monitoring & Recording (702) Counter module (703) Pricing module (704)  Other components integrated with the web research agent module (106); Platform management module (101) 20 o Intent detection module (102) Event trigger module (103) Business Process Flow Module Platform channel module (107) o Input security layer (201) 25 o API gateway (202) o Identity and access management (203) o Great language model (301) Planning agents (302) Evaluation agents and feedback module (303) 30 Predefined task agents (304) o Vehicle caller (305) that Adapter (401) Smart operation module (501) o Unstructured data (503) 35 45 o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703)  Other components integrated with the platform channel module (107); 5 Platform management module (101) o Intent detection module (102) Event trigger module (103) Business process flow module (104) Web research agent module (106) 10 o Input security layer (201) o API gateway (202) o Identity and access management (203) o Great language model (301) Planning agents (302) 15 Evaluation agents and feedback module (303) Predefined task agents (304) o Vehicle caller (305) that Adapter (401) Business support systems (402) 20 Operation support systems (403) that Connector (404) o External software applications (405) Fixed and mobile network devices (406) Other Artificial Intelligence Application Tools 25 databases (408) External cloud services (409) those Servers (410) Smart operation module (501) Data lake and repository (701) 30 o Monitoring & Recording (702)  Other components integrated with the input security layer (201); Platform management module (101) o Intent detection module (102) 35 46 Event trigger module (103) Business process flow module (104) Timer module (105) Web research agent module (106) Platform channel module (107) 5 o API gateway (202) o Identity and access management (203) o Great language model (301) Planning agents (302) Evaluation agents and feedback module (303) 10 Predefined task agents (304) o Data lake and repository (701) o Monitoring & Recording (702)  Other components integrated with the API gateway (202); 15 Platform management module (101) o Intent detection module (102) Event trigger module (103) Business Process Flow Module Timer module (105) 20 Web research agent module (106) Platform channel module (107) o Input security layer (201) o Identity and access management (203) o Great language model (301) 25 Planning agents (302) Evaluation agents and feedback module (303) Predefined task agents (304) o Vehicle caller (305) that Adapter (401) 30 Business support systems (402) Operation support systems (403) that Connector (404) o External software applications (405) Fixed and mobile network devices (406) 35 47 Other Artificial Intelligence Application Tools databases (408) External cloud services (409) those Servers (410) Smart operation module (501) 5 o Structured data (502) o Unstructured data (503) o Knowledge base (504) Synthetic data (505) Business process pool module (601) 10 o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703) Pricing module (704)  Other components integrated with identity and access management (203); 15 Platform management module (101) o Intent detection module (102) Event trigger module (103) Business process flow module (104) Timer module (105) 20 Web research agent module (106) Platform channel module (107) o Input security layer (201) o API gateway (202) o Great language model (301) 25 Planning agents (302) Evaluation agents and feedback module (303) Predefined task agents (304) o Vehicle caller (305) that Adapter (401) 30 Business support systems (402) Operation support systems (403) that Connector (404) o External software applications (405) Fixed and mobile network devices (406) 35 48 Other artificial intelligence application tools (407) databases (408) External cloud services (409) those Servers (410) Smart operation module (501) 5 o Structured data (502) o Unstructured data (503) o Knowledge base (504) Synthetic data (505) Business process pool module (601) 10 o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703) Pricing module (704) Other components integrated with the large language model (301); 15 Platform management module (101) o Intent detection module (102) Event trigger module (103) Business process flow module (104) Timer module (105) 20 Web research agent module (106) Platform channel module (107) o Input security layer (201) o API gateway (202) Identity and access management (203) 25 Planning agents (302) Evaluation agents and feedback module (303) Predefined task agents (304) o Vehicle caller (305) o Unstructured data (503) 30 o Knowledge base (504) Synthetic data (505) Business process pool module (601) o Data lake and repository (701) 49 o Monitoring & Recording (702) Counter module (703)  Other components integrated with planning agents (302); Platform management module (101) o Intent detection module (102) 5 Event trigger module (103) Business process flow module (104) Timer module (105) Web research agent module (106) Platform channel module (107) 10 o Input security layer (201) o API gateway (202) o Identity and access management (203) o Great language model (301) Evaluation agents and feedback module (303) 15 Predefined task agents (304) Business process pool module (601) o Data lake and repository (701) o Monitoring & Recording (702)  The other 20 integrated with the evaluation agents and feedback module (303) components; Platform management module (101) Web research agent module (106) Platform channel module (107) o Input security layer (201) 25 o API gateway (202) o Identity and access management (203) o Great language model (301) Planning agents (302) Predefined task agents (304) 30 Business process pool module (601) o Data lake and repository (701) o Monitoring & Recording (702)  Other components integrated with predefined mission agents (304); Platform management module (101) 35 50 Web research agent module (106) Platform channel module (107) o Input security layer (201) o API gateway (202) o Identity and access management (203) 5 o Great language model (301) Planning agents (302) Evaluation agents and feedback module (303) o Vehicle caller (305) Data lake and repository (701) 10 o Monitoring & Recording (702)  Other components integrated with the vehicle caller (305); Platform management module (101) Web research agent module (106) Platform channel module (107) 15 o API gateway (202) o Identity and access management (203) o Great language model (301) Predefined task agents (304) that Adapter (401) 20 o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703)  Other components integrated with the adapter (401); Platform management module (101) 25 Web research agent module (106) Platform channel module (107) o API gateway (202) o Identity and access management (203) Vehicle caller (305) 30 Business support systems (402) Operation support systems (403) o External software applications (405) Fixed and mobile network devices (406) Other Artificial Intelligence Application Tools 35 51 databases (408) External cloud services (409) those Servers (410) Smart operation module (501) o Structured data (502) 5 o Unstructured data (503) o Knowledge base (504) o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703) 10 Pricing module (704)  Other components integrated with business support systems (402); Platform management module (101) Event trigger module (103) Platform channel module (107) 15 o API gateway (202) o Identity and access management (203) that Adapter (401) that Connector (404) o Structured data (502) 20 o Unstructured data (503) o Knowledge base (504) o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703) 25  Other components integrated with operational support systems (403); Platform management module (101) Event trigger module (103) Platform channel module (107) o API gateway (202) 30 o Identity and access management (203) that Adapter (401) that Connector (404) o Structured data (502) o Unstructured data (503) 35 52 o Knowledge base (504) o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703)  Other components integrated with connector (404); 5 Platform management module (101) Platform channel module (107) o API gateway (202) o Identity and access management (203) that Adapter (401) 10 Business support systems (402) Operation support systems (403) o External software applications (405) Fixed and mobile network devices (406) Other artificial intelligence application tools (407) 15 databases (408) External cloud services (409) those Servers (410) Smart operation module (501) o Structured data (502) 20 o Unstructured data (503) o Knowledge base (504) o Data lake and repository (701) o Monitoring & Recording (702)  Other components integrated with external software applications (405); 25 Platform management module (101) Event trigger module (103) Platform channel module (107) o API gateway (202) Identity and access management (203) 30 that Adapter (401) that Connector (404) Smart operation module (501) o Structured data (502) o Unstructured data (503) 35 53 o Knowledge base (504) o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703)  Other components integrated with fixed and mobile network devices (406); 5 Platform management module (101) Event trigger module (103) Platform channel module (107) o API gateway (202) o Identity and access management (203) 10 that Adapter (401) that Connector (404) Smart operation module (501) o Structured data (502) o Unstructured data (503) 15 o Knowledge base (504) o Data lake and repository (701) o Monitoring & Recording (702) Counter module (703)  Other components integrated with other artificial intelligence application tools (407); 20 Platform management module (101) Event trigger module (103) Platform channel module (107) o API gateway (202) Identity and access management (203) 25 that Adapter (401) that Connector (404) Smart operation module (501) o Structured data (502) o Unstructured data (503) 30 o Knowledge base (504) o Data lake and repository (701) Monitoring and Recording Counter module (703)  Other components integrated with databases (408); 35 54 Platform management module (101) Event trigger module (103) Platform channel module (107) o API gateway (202) o Identity and access management (203) 5 that Adapter (401) that Connector (404) Smart operation module (501) o Structured data (502) o Unstructured data (503) 10 o Knowledge base (504) o Data lake and repository (701) Monitoring and Recording Counter module (703) o Pricing module (704) 15  Other components integrated with external cloud services (409); Platform management module (101) Event trigger module (103) Platform channel module (107) o API gateway (202) 20 o Identity and access management (203) that Adapter (401) that Connector (404) Smart operation module (501) o Structured data (502) 25 o Unstructured data (503) o Knowledge base (504) o Data lake and repository (701) Monitoring and Recording Counter module (703) 30  Other components integrated with the servers (410); Platform management module (101) Platform channel module (107) o API gateway (202) Identity and access management (203) 35 55 that Adapter (401) that Connector (404) Smart operation module (501) o Structured data (502) o Unstructured data (503) 5 o Knowledge base (504) Synthetic data (505) o Data lake and repository (701) Monitoring and Recording Counter module (703) 10 Pricing module (704)  Other components integrated with the intelligent operation module (501); Platform management module (101) Event trigger module (103) Business Process Flow Module 15 Timer module (105) Web research agent module (106) Platform channel module (107) o API gateway (202) o Identity and access management (203) 20 that Adapter (401) that Connector (404) o External software applications (405) Fixed and mobile network devices (406) Other Artificial Intelligence Application Tools 25 databases (408) External cloud services (409) those Servers (410) o Structured data (502) o Unstructured data (503) 30 o Knowledge base (504) Synthetic data (505) o Data lake and repository (701) Monitoring and Recording o Counter module (703) 35 56 Pricing module (704)  Other components integrated with structured data (502); Platform management module (101) o API gateway (202) o Identity and access management (203) 5 that Adapter (401) Business support systems (402) Operation support systems (403) that Connector (404) External software applications (405) 10 Fixed and mobile network devices (406) Other Artificial Intelligence Application Tools databases (408) External cloud services (409) those Servers (410) 15 Smart operation module (501) o Knowledge base (504) Synthetic data (505) o Data lake and repository (701) Monitoring and Recording 20 Counter module (703) Pricing module (704)  Other components integrated with unstructured data (503); Platform management module (101) o Intent detection module (102) 25 Web research agent module (106) o API gateway (202) o Identity and access management (203) o Great language model (301) that Adapter (401) 30 Business support systems (402) Operation support systems (403) that Connector (404) o External software applications (405) Fixed and mobile network devices (406) 35 57 Other Artificial Intelligence Application Tools databases (408) External cloud services (409) those Servers (410) Smart operation module (501) 5 o Knowledge base (504) Synthetic data (505) Business process pool module (601) o Data lake and repository (701) Monitoring and Recording 10 Counter module (703) Pricing module (704)  Other components integrated with the knowledge base (504); Platform management module (101) o Intent detection module (102) 15 o API gateway (202) o Identity and access management (203) o Great language model (301) that Adapter (401) Business support systems (402) 20 Operation support systems (403) that Connector (404) o External software applications (405) Fixed and mobile network devices (406) Other Artificial Intelligence Application Tools 25 databases (408) External cloud services (409) those Servers (410) Smart operation module (501) o Structured data (502) 30 o Unstructured data (503) Synthetic data (505) Business process pool module (601) o Data lake and repository (701) Monitoring and Recording 35 58 Counter module (703) Pricing module (704)  Other components integrated with synthetic data (505); Platform management module (101) o API gateway (202) 5 o Identity and access management (203) o Great language model (301) those Servers (410) Smart operation module (501) o Structured data (502) 10 o Unstructured data (503) o Knowledge base (504) Monitoring and Recording  Other components integrated with the business process pool module (601); Platform management module (101) 15 Business Process Flow Module o API gateway (202) o Identity and access management (203) o Great language model (301) Planning agents (302) 20 Evaluation agents and feedback module (303) o Unstructured data (503) o Knowledge base (504) o Data lake and repository (701) Monitoring and Recording 25  Data lake and other components integrated with the data repository (701); Platform management module (101) o Intent detection module (102) Event trigger module (103) Business Process Flow Module 30 Timer module (105) Web research agent module (106) Platform channel module (107) o Input security layer (201) o API gateway (202) 35 59 o Identity and access management (203) o Great language model (301) Planning agents (302) Evaluation agents and feedback module (303) Predefined task agents (304) 5 o Vehicle caller (305) that Adapter (401) Business support systems (402) Operation support systems (403) Connector (404) 10 o External software applications (405) Fixed and mobile network devices (406) Other Artificial Intelligence Application Tools databases (408) External cloud services (409) 15 those Servers (410) Smart operation module (501) o Structured data (502) o Unstructured data (503) o Knowledge base (504) 20 Synthetic data (505) Business process pool module (601) Monitoring and Recording Counter module (703) o Pricing module (704) 25  Other components integrated with Monitoring & Recording (702); Platform management module (101) o Intent detection module (102) Event trigger module (103) Business Process Flow Module 30 Timer module (105) Web research agent module (106) Platform channel module (107) o Input security layer (201) o API gateway (202) 35 60 o Identity and access management (203) o Great language model (301) Planning agents (302) Evaluation agents and feedback module (303) Predefined task agents (304) 5 o Vehicle caller (305) that Adapter (401) Business support systems (402) Operation support systems (403) Connector (404) 10 o External software applications (405) Fixed and mobile network devices (406) Other Artificial Intelligence Application Tools databases (408) External cloud services (409) 15 those Servers (410) Smart operation module (501) o Structured data (502) o Unstructured data (503) o Knowledge base (504) 20 Synthetic data (505) Business process pool module (601) o Data lake and repository (701) Counter module (703) o Pricing module (704) 25  Other components integrated with the counter module (703); Platform management module (101) o Intent detection module (102) Event trigger module (103) Business Process Flow Module 30 Timer module (105) Web research agent module (106) o API gateway (202) o Identity and access management (203) o Great language model (301) 35 61 o Vehicle caller (305) that Adapter (401) Business support systems (402) Operation support systems (403) External software applications (405) 5 Fixed and mobile network devices (406) Other Artificial Intelligence Application Tools databases (408) External cloud services (409) those Servers (410) 10 Smart operation module (501) o Structured data (502) o Unstructured data (503) o Knowledge base (504) o Data lake and repository (701) 15 Monitoring and Recording Pricing module (704)  Other components integrated with the billing module (704); Platform management module (101) Event trigger module (103) 20 Timer module (105) o API gateway (202) o Identity and access management (203) that Adapter (401) those databases (408) 25 those Servers (410) Smart operation module (501) o Structured data (502) o Unstructured data (503) o Knowledge base (504) 30 o Data lake and repository (701) o Monitoring and Recording (702) Counter module (703) 62 A1.4. Developer (802), for role and responsibility-based platform management, each of the platform Development to be integrated with the identity and access management (203) module in the component. does. The platform is flexible, continuously improvable, and supports different telecommunications operators. (804) or the changing structures of Service Providers or Businesses in different sectors 5 It should be adaptable and configurable to meet diverse needs. It is necessary. After the platform is activated, the admin (801) can send the following instructions as needed. The developer will configure the platform components. (802) The platform can be configured To achieve this, role and responsibility-based enhancements are needed in the following platform components: These roles are admin (801) and 10 who perform basic operations on the system. is the operator (804). In this way, the developer (802) defines an authorization level for each role; the admin (801) Different configuration processes are carried out according to authorization levels and the operator (804) completes the improvements that will enable it to perform the operation. Authentication and authorization of individuals accessing the platform is part of identity and access management. This is done thanks to module (203). 15  Platform management module (101), o Admin (801): Operator (804):  Intent detection module (102), o Admin (801): 20 Operator (804):  Event trigger module (103), o Admin (801): Operator (804):  Business Process Flow Module, 25 o Admin (801): Operator (804):  Timer module (105), o Admin (801): Operator (804): 30  Web research agent module (106) o Admin (801): Operator (804):  Input security layer (201) o Admin (801): 35 63 Operator (804):  API gateway (202) o Admin (801): Operator (804):  Identity and access management (203) 5 o Admin (801): Operator (804):  Large language model (301) o Admin (801): Operator (804): 10  Planning agents (302) o Admin (801): Operator (804):  Evaluation agents and feedback module (303) o Admin (801): 15 Operator (804):  Predefined mission agents (304) o Admin (801): Operator (804):  Vehicle caller (305) 20 o Admin (801): Operator (804):  Adapter (401) o Admin (801): Operator (804): 25  Business support systems (402) o Admin (801): Operator (804):  Operation support systems (403) o Admin (801): 30 Operator (804):  Connector (404) o Admin (801): Operator (804):  External software applications (405) 35 64 o Admin (801): Operator (804):  Fixed and mobile network devices (406) o Admin (801): Operator (804): 5  Other artificial intelligence application tools (407) o Admin (801): Operator (804):  Databases (408) o Admin (801): 10 Operator (804):  External cloud services (409) o Admin (801): Operator (804):  Servers (410) 15 o Admin (801): Operator (804):  Smart operation module (501) o Admin (801): Operator (804): 20  Structured data (502) o Admin (801): Operator (804):  Unstructured data (503) o Admin (801): 25 Operator (804):  Knowledge base (504) o Admin (801): Operator (804):  Synthetic data (505) 30 o Admin (801): Operator (804):  Business process pool module (601) o Admin (801): Operator (804): 35 65  Data lake and repository (701) o Admin (801): Operator (804):  Monitoring & Recording (702) o Admin (801): 5 Operator (804):  Counter module (703) o Admin (801): Operator (804):  Pricing module (704) 10 o Admin (801): Operator (804): A2. Platform Configuration Step: The platform is owned by a telecommunications operator (804) or service provider or any After it is made available for use by the business, but only before the business process is implemented, admin 15 (801) Configure the platform according to business processes in the following platform components;  Platform management module (101)  Intent detection module (102),  Event trigger module (103),  Business Process Flow Module, 20  Timer module (105),  Web research agent module (106)  Input security layer (201)  API gateway (202)  Identity and access management (203) 25  Large language model (301)  Planning agents (302)  Evaluation agents and feedback module (303)  Predefined mission agents (304)  Vehicle caller (305): Model Context Protocol 30 for each vehicle to be added The tool caller (305) has been developed for updating. Integrations have also been made. It is completed.  Adapter (401)  Business Support Systems 66  Operation support systems (403)  Connector (404)  External software applications (405)  Fixed and mobile network devices (406)  Other artificial intelligence application tools (407) 5  Databases (408)  External cloud services (409)  Servers (410)  Smart operation module (501)  Structured data (502) 10  Unstructured data (503)  Knowledge base (504)  Synthetic data (505)  Business process pool module (601)  Data lake and repository (701) 15  Monitoring & Recording (702)  Counter module (703)  Pricing module (704) Each new business process, new task, new subtask, and new tool can be added to the platform, totaling 20. When this happens, the relevant modules on the platform will be configured. A3. Platform Operation Step: The operational steps of the platform, once it is ready for use and commissioned, are as follows: It is like this; A3.1. The system is triggered by human or external stimuli; 25  Admin (801), customer (803), operator (804) platform and platform channel module (107) communicates with the platform.  Smart operation module (501) or any external software application or Platform channel module, platform-integrated trigger such as an IoT device. (107) sends a signal. 30  The timer module (105) was previously created by admin (801) on the platform. Scheduled tasks trigger the platform. 67  The web research agent module (106) was previously accessed by admin (801) on the platform. Tasks executed and completed through the platform trigger it. A3.2. External triggers to the platform are forwarded to the relevant module for evaluation;  Requests from the admin (801), customer (803), or operator (804), 5 with intent detection module (102) transmitted via platform channel module (107) By analyzing, the intention of the admin (801), the customer (803) or the operator (804) The identification is made. The intention, which matches a meaningful and predefined pattern, is found in a specific text. It is transmitted to the planning agents (302) in the format (with a predefined request). For example Customer (803) “I received a higher bill than I expected” and “The charge is every Both complaints, "More than usual," and "Customer billing amount 10 It is transformed into a request in the form of, "They contacted us due to a complaint."  From the intelligent operation module (501) or any external software via platform channel module (107) from application or IoT device The transmitted signals are analyzed by the event trigger module (103). Analysis As a result, the event trigger module (103) is in a predefined format 15 It transmits a prepared text input (request) to the planning agents (302). For example According to information from the intelligent operation module (501), the OLT network device is above If an anomaly indicating saturation in directional traffic is detected, the event The trigger module (103) detects this on the uplink card of the OLT device named “X” "A saturation level of x% has occurred in the traffic." This was the kind of request made on page 20. communicates to planning agents (302).  The timer module (105) was previously created by admin (801) on the platform. Scheduled tasks via the platform channel module (107) This can trigger the platform. For example, the most suitable campaign will be applied every night. Identifying customer (803) segments and customer (803) campaign 25 once the matching process is complete and the campaign is implemented, with a specific request. forwarding the segmented customer (803) list to planning agents (302) For example.  The web research agent module (106) was previously accessed by admin (801) on the platform. Tasks executed and completed via platform channel module (107) 30 It can trigger the platform through. For example, dropped onto a specific web application. Notification regarding breaking news announcement by web research agent module (106) It transmits to planning agents (302) with a specific request format. 68 A3.3. Planning agents (302) from intent detection module (102), event trigger from module (103), timer module (105), web search agent module (106) Creates tasks according to appropriate requests with incoming triggers. A3.4. Planning agents (302), work located in the business process pool module (601) It identifies the most suitable process(es) from among the available processes, taking the task as a reference, and this work is 5 Taking into account the suggestions, recommendations, and guidance in the processes, the task is delegated to subordinate tasks. It divides and determines the order of subtasks. Each separated subtask is pre-selected. It matches the defined task agents (304) and assigns them predefined tasks to perform. Assigns sub-tasks to task agents (304). A3.5. Planning agents (302) Continuously monitor the progress of each sub-task, 10 According to feedback from the evaluation agents and the feedback module (303) It makes revisions by updating at each stage. The feedback process, in terms of its integrity within the process, is expressed as follows: It is carried out as follows; A3.5.1 Initial Configuration – admin (801), business process pool module (601) 15 potential triggers, tasks, subtasks, task agents, and tools for each business process. It defines and weights the mappings. These definitions are for planning agents (302). It is used as an initial policy. A3.5.2 Work Process and Monitoring – Planning agents (302) when each trigger occurs selects the appropriate work process, breaks it down into subtasks, and uses appropriate task agents and tools. 20 It matches. The evaluation agents and feedback module (303) then perform these matches. It evaluates its performance using defined metrics; for example, if the processing time is long. It provides feedback indicating that the task agent or vehicle selection was unsuccessful. A3.5.3 IGBPÖ Data Collection – admin (801) or authorized human users, different By comparing the matching results, they can choose which one is better. 25 It reports that these preferences are mostly in the form of pairwise comparisons. Collected Comparisons create a data set. A3.5.4 Reward Model and Reinforcement Learning – With collected comparison data The reward model is trained; the model assigns a numerical reward value to each matching chain. This reward model evaluates the matching policy of planning agents (302). It is used in updating with an algorithm like PPO. Thus, planning agents (302), metric-based feedback (from Assessment Agents) and human It optimizes its policy by considering the reward that its preferences represent. A3.5.5 Continuous Improvement – ​​Evaluation and Feedback provided by Agents Performance metrics can also be used as part of the IGBPÖ cycle. 35 69 For example, the duration of a subtask, error rate or customer (803) satisfaction score reward These can be used as features of the model. Thus, the reward model can be supported by both numerical data. It is fed and calibrated by human preferences. In the third stage, admin (801) new It collects preference data and updates the reward model; reinforcement is applied in the fourth stage. The policy is updated as it learns; this process is repeated periodically. In this way, 5 The business process pool module (601) which will be used by planning agents (302) The matching suggestion is “business process → task → subtask → task agent → tool” consistently. It is being improved. A3.6. Planning agents (302) will execute the sub-tasks matched by 10 in advance. Defined task agents (304) perform the function of the relevant vehicle via the vehicle caller (305). It operates. These tools are business support systems (402), operations depending on the agent's task. support systems (403), external software applications (405), fixed and mobile network devices (406), Other Artificial Intelligence Application Tools, databases (408), external cloud There can be many types of tools, such as services (409), servers (410). 15 A3.7. All transactions, logs and events are stored in the data lake and repository (701). A3.8. Tokens for completed transactions are counted using the counter module (703) and charged accordingly. With module (704) the fee per unit transaction is calculated and invoiced to the customer (803) The process is completed. Payment methods include prepaid options like credit cards and mobile payments, as well as adding to a bill. Post-payment systems are integrated into the platform. 20 A4. Operational Step of the Platform: The platform is not owned by a telecommunications operator (804) or service provider or any After it is made available for use by the business, the operator (804), admin (801) and developer (802) They perform operational procedures in case of errors and malfunctions in each component. 25 operator (804) Level 1, admin (801) Level 2 and developer (802) Level 3 type It meets the operational needs.  Platform management module (101)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801) 30  Level 3 Operation: Developer (802)  Intent detection module (102)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802) 35  Event trigger module (103)  Level 1 Operation: Operator (804) 70  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Business process flow module (104)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801) 5  Level 3 Operation: Developer (802)  Timer module (105)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802) 10  Web research agent module (106)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Input security layer (201) 15  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  API gateway (202)  Level 1 Operation: Operator (804) 20  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Identity and access management (203)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801) 25  Level 3 Operation: Developer (802)  Large language model (301)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802) 30  Planning agents (302)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Evaluation agents and feedback module (303) 35  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Predefined mission agents (304)  Level 1 Operation: Operator (804) 40  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Vehicle caller (305)  Level 1 Operation: Operator (804) 71  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Adapter (401)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801) 5  Level 3 Operation: Developer (802)  Business support systems (402)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802) 10  Operation support systems (403)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Connector (404) 15  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  External software applications (405)  Level 1 Operation: Operator (804) 20  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Fixed and mobile network devices (406)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801) 25  Level 3 Operation: Developer (802)  Other Artificial Intelligence Application Tools  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802) 30  Databases (408)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  External cloud services (409) 35  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Servers (410)  Level 1 Operation: Operator (804) 40  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Smart operation module (501)  Level 1 Operation: Operator (804) 72  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Structured data (502)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801) 5  Level 3 Operation: Developer (802)  Unstructured data (503)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802) 10  Knowledge base (504)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Synthetic data (505) 15  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Business process pool module (601)  Level 1 Operation: Operator (804) 20  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802)  Data lake and repository (701)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801) 25  Level 3 Operation: Developer (802)  Monitoring & Recording (702)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802) 30  Counter module (703)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802) 35  Pricing module (704)  Level 1 Operation: Operator (804)  Level 2 Operation: Admin (801)  Level 3 Operation: Developer (802) 40 73 The platform can be a human (admin (801), customer (803), operator (804)) or a system (external). platform channel by applications, IoT devices, timers, web agents) It is triggered via module (107). All incoming triggers are routed to the relevant analysis modules:  Intent Detection Module: Requests from people are analyzed and standardized into 5 requests. It is transformed.  Event trigger module (103): System signals (e.g., in a network device (anomaly) is converted into text prompts.  Timer and Web Research Agent: Pre-scheduled tasks, ready It triggers the platform with requests. 10 Planning agents (302) perform appropriate tasks according to the requests from the modules above. It creates. Planning agents (302) select the most suitable business processes from the Business Process Pool, their tasks It divides the task into subtasks and maps these subtasks to predefined task agents (304). Planning agents (302), each step evaluation agents and feedback module (303) 15 By monitoring through it, it continuously improves its tasks. Predefined task agents (304), via vehicle caller (305) related systems (BSS, OSS, network devices, software, data It runs databases (408), AI tools, cloud services, etc.). All actions and logs are stored in the data lake. and is stored in its warehouse (701). Token counting for each transaction is done with the counter module (703), then the fee is 20 It is calculated with module (704). The payment infrastructure is both prepaid and postpaid. It is integrated with the systems.

Claims

74 REQUESTS 1. Operators (804) and services operating in the telecommunications sector providers, technology, marketing, sales, finance, procurement, law, regulation, processes carried out in different business units such as human resources and customer (803) relations without human intervention or with human support through artificial intelligence agents 5 It is an artificial intelligence agent platform that performs the following function;  The system administrator must have access to all modules, agents, and the knowledge base of the platform. (504), data sources, business process pool and tool context protocols Platform management that enables configuration, management, and monitoring. module (101), 10  Analyzes incoming triggers, selects the appropriate work process, and subdivides the task. feedback from agents that divide, assign subtasks, and evaluate the process. Planning agent constantly revising according to notifications (302),  Transmits user or system-generated triggers to the platform, in audio, text, or otherwise. Platform channel 15 provides bidirectional communication via data-based protocols. module (107),  Analyze natural language inputs provided by the user or artificial intelligence by determining the user's intention from the input, the planning agent (302) Intent detection module transmitting in standard format (102),  Monitors technical, operational, or network-based events in the system, changes, or 20 event trigger that activates the planning agent (302) when it detects an anomaly module (103),  Maintaining the sequence and conditional relationships between multiple business processes, specific events or when the intention is realized, the planning agents of the relevant business process (302) Business process flow module (104) that enables implementation by 25  certain business processes at times defined by admin (801) or enable the planning agent (302) to start automatically in periods triggering timer module (105),  Monitoring external data sources or the web environment, in case of a defined development or Web research agent module 30, which triggers the platform when a change is detected. (106),  repetitive sub-tasks in telecommunications business processes predefined task agents that carry out tasks without requiring intervention (304), 75  System, network device required by predefined task agents (304), contextually necessary; enables interaction with databases or external services. Vehicle caller (305) that calls vehicles via API or private protocols.  by planning agents (302) and predefined task agents (304) The results of the processes carried out are evaluated in terms of performance, time, resource utilization, and 5. evaluation agencies that assess users using criteria such as user satisfaction and feedback module (303),  When dividing the tasks of the planning agents (302) in the business process into sub-tasks, prioritizing, sequencing, and predefined task agents for subtasks (304) is accepted as a reference in order to make correct decisions when matching with work 10 Workers who carry out the advice, suggestions and directives of the admin (801) regarding the process. It includes the process pool module (601).

2. According to claim 1, it is an artificial intelligence agent platform whose feature is reinforcement learning. using algorithms, the decision policies of planning agents (302) 15 It includes updating evaluation agents and feedback module (303).

3. It is an artificial intelligence agent platform according to claim 1, and its feature is;  intent detection module (102), event trigger module (103), business processes flow module (104), timer module (105), web search agent 20 Configuring module (106),  intelligent operation module (501), structured data (502), unstructured data (503), knowledge base (504), synthetic data (505), predefined Managing the task agents (304), the adapter (401) and the vehicle caller (305),  Management 25 that manages the business processes in the business process pool module (601) It includes module (101).

4. It is an artificial intelligence agent platform according to claim 1, and its feature is; from the admin (801), from customers (803), operator (804) or any person in any role or artificial intelligence derived from natural language input; in the specified requests or 30 Planning by understanding the intentions behind the complaints and detailing them in a specific format. It includes the intent detection module (102) which transmits to its agent (302). 76 5. It is an artificial intelligence agent platform according to claim 1, and its feature is that the admin's (801) business process The implementation time setting is set on the timer module (105) with the work process. It does this by matching time information.

6. It is an artificial intelligence agent platform according to claim 1, and its feature is; continuous or periodic 5 It includes a web search agent module (106) that scans external sources.

7. According to Claim 1, it is an artificial intelligence agent platform, and its characteristic is; the aforementioned platform channel module (107);  an AI call center that communicates two-way with the customer (803) his assistant,  A human call center assistant communicating two-way with the customer (803) tools and system,  Two-way communication with the customer (803) and platform services 15 through an integrated mobile application,  providing two-way communication with the customer (803) and services with the platform through an integrated web application,  Special platform that enables two-way communication with the customer (803) and is integrated with the platform services, 20  Application software that integrates with the platform and establishes bidirectional communication. services,  IoT or closed system that is integrated with the platform and communicates bidirectionally. services between devices  25 requiring platform integration and near real-time data flow It includes system messages.

8. According to Claim 1, it is an artificial intelligence agent platform whose feature is instant messaging and The platform supports chat channels and includes a channel module (107).

9. It is an artificial intelligence agent platform according to Claim 1, and its feature is high data carrying capacity and a platform with infrastructure that supports scenarios requiring low latency It contains channel module (107). 77 10. According to claim 1, it is an artificial intelligence agent platform and its features are; admin (801), client (803) and platform channel that enables the operator (804) to communicate with the platform. It includes module (107).

11. It is an artificial intelligence agent platform according to claim 1, and its feature is; 5  When intent is detected from the intent detection module (102),  An event occurs, its status changes, and it is observed from the event trigger module (103). when a change is detected as a result,  From the business process flow module (104), when it is the turn of the business process in the workflow 10 when detected,  The operating time is set or adaptive from the timer module (105). when it is determined that a timed business process has been triggered,  In the web research agent module (106) the result of the research activity Task 15 is triggered when reached, and the triggering result provides meaningful information. It includes the planning agents (302) that make up.

12. It is an artificial intelligence agent platform according to claim 1, and its feature is; it receives from triggers. information and business process directions in the business process pool module (601) analyzes and identifies tasks that will ensure the fulfillment of business processes, sub-20 structuring into tasks and assigning to predefined task agents (304) It includes planning agents (302) that enable the system to work by distributing them.

13. It is an artificial intelligence agent platform according to Claim 1, and its characteristic is the sequencing of sub-tasks. and the priority, dependency and timing relationships between these sub-tasks 25 It includes planning agents (302) that generate flow by evaluating.

14. It is an artificial intelligence agent platform according to Claim 1, and its feature is; planning agents. (302) independently scalable, independently versionable and other It is a service that can operate independently of agents. 30 15. It is an artificial intelligence agent platform according to Claim 1, and its feature is that it performs tasks holistically. Planning agents (302) are a central coordination layer that is controlled It includes. 35 78 16. It is an artificial intelligence agent platform according to Claim 1, and its feature is; predefined task agents (304), evaluation agents and feedback module (303) asynchronous It includes planning agents (302) who communicate as 17. It is an artificial intelligence agent platform according to Claim 1, and its feature is; planned tasks and 5 to monitor the quality of the output produced, generate meaningful feedback and the process assessment, which is an integrated intelligence component structured for the purpose of guidance. It includes agents and feedback module (303).

18. According to claim 1, it is an artificial intelligence agent platform, and its feature is; critical analysis of the outputs. By conducting analyses, offering improvement suggestions, and re-evaluating tasks. providing feedback to the planning agent (302) for its structuring It includes evaluation agents and feedback module (303). According to claim 19, it is an artificial intelligence agent platform and its feature is; vehicle summoner (305) 15 by running the function of the relevant tool via the developers (802) as needed by developing and making available on the platform predefined task agents (304) It includes.

20. It is an artificial intelligence agent platform according to Claim 1, and its feature is to perform specific tasks. structured to be operationally focused and capable of taking direct action, planning Sub-tasks determined by the agents (302) during the planning phase It includes predefined task agents (304) that work to accomplish this.

21. It is an artificial intelligence agent platform according to Claim 1, and its feature is; 25 for specific task types. Creating optimized, highly efficient, action-oriented modules, pre-determined outputs that enable the system to produce deterministic and repeatable outputs It includes defined task agents (304). According to claim 1, it is an artificial intelligence agent platform and its feature is; planning agents (302), 30 planning that works integrated with the evaluation agents and feedback module (303) receiving a mission order from agents (302), carrying out the operation and the result transmitting to evaluation and feedback agents or to a central log system It includes predefined mission agents (304). 35 23. According to Claim 1, it is an artificial intelligence agent platform, and its characteristic is; model context. tool caller (305) that runs the protocol on the cluster of servers (410) It includes. 79 24. It is an artificial intelligence agent platform according to Claim 1, and its feature is that the task agents are specific functions defined on tools or application programming interfaces tool caller (305) which provides the executable context necessary for calling It includes. 5 According to Claim 1, it is an artificial intelligence agent platform characterized by high availability and Microservice-based, horizontally integrated, with context processing performance in mind. It includes a tool caller (305) distributed in a scalable structure.

Images