NIRNAYA: An AI-Driven Intelligent Support Ticketing and Decision System

NIRNAYA: An AI-Driven Intelligent Support Ticketing and Decision System

Mrs. Bharathi K.

Department of Computer Science and Engineering Acharya Institute of Technology, Bengaluru, India

Malay Shekhar, Kushagra Kaushik, Atul Raj, K. Umar Sadiq

Department of Computer Science and Engineering Acharya Institute of Technology, Bengaluru, India

Abstract – Internal IT support is an essential part of the daily operation of modern organizations. However, a consider- able portion of the support process still depends on manually reading requests, identifying their type and urgency, selecting the appropriate department, and assigning them to suitable support personnel. These activities can become time-consuming when ticket volumes increase and may also result in inconsistent decisions for similar requests.

The proposed system, NIRNAYA, is an AI-driven internal IT

Service Management (ITSM) platform designed to assist support teams during the decision-intensive stages of ticket handling. The system uses natural language processing to analyze employee requests and generate recommendations for ticket classication, priority, department routing, and agent assignment. It also pro- vides ticket summarization and initial resolution assistance. The platform combines a React-based user interface, Node.js/Next.js application services, PostgreSQL for persistent data storage, AI-based language processing services, and WebSocket-based real-time communication. SLA monitoring, notications, ticket history, and operational analytics are integrated into the same workow.

A key aspect of NIRNAYA is its human-in-the-loop design. AI-

generated recommendations are presented to support personnel for validation rather than being treated as nal decisions. This allows the proposed system to reduce repetitive triage work while preserving human control over ticket handling. The proposed ar- chitecture incorporates AI-assisted ticket processing, structured data management, SLA monitoring, real-time communication, and operational analytics within a unied IT support workow. NIRNAYA is presented as a system design and prototype; quantitative evaluation of prediction accuracy and operational performance is identied as an area for future work.

Index TermsIT Service Management, intelligent help desk,

articial intelligence, natural language processing, ticket classi- cation, priority prediction, intelligent routing, SLA monitoring, decision support.

1. Introduction

   Information technology support plays an important role in keeping organizations productive. Employees may encounter a wide range of technical problems, including VPN failures, Wi-Fi connectivity issues, email problems, password and access requests, hardware failures, software errors, database issues, and security-related incidents. In a conventional support environment, an employee submits a request and a support coordinator or agent manually interprets the description before deciding how the ticket should be handled.

   The initial interpretation of a ticket often involves several decisions. Support personnel must determine what type of problem has been reported, how urgent it is, which department

   should handle it, and which available agent is most appropri- ate. Only after these decisions have been made can the ticket move effectively toward resolution.

   Previous research has investigated automated approaches for classifying customer and IT support tickets. Truss and Bo¨hm studied AI-based customer support ticket classication using AutoML and examined the applicability of automated machine- learning approaches to support-ticket classication [?].

   Similarly, Paramesh et al. investigated the classication of unstructured IT service-desk tickets using ensemble learning techniques. Their work examined bagging, boosting, and vot- ing approaches for improving ticket classication [?].

   Al-Hawari and Barham developed a machine-learning-based help desk system for IT service management in which ticket classication is used to associate tickets with appropriate services [?]. Benitez Pereira et al. also investigated machine- learning classication of IT support tickets and developed a prototype for assisting IT analysts [?].

   Consequently, the challenge in IT support is not limited to storing tickets and tracking their status. A signicant amount of effort is spent during the early stages of the workow, particularly when requests are received in free-form natural language. Similar problems may also be classied differently by different support personnel, resulting in variation in routing and prioritization.

   The proposed system, NIRNAYA, is designed to assist with this decision-making process. The system analyzes an employees request and generates structured recommendations that can be reviewed by support personnel. These recommen- dations include the likely category of the issue, its priority, the responsible department, and a suitable support agent. The system can additionally generate a concise summary and provide initial troubleshooting guidance.

   The proposed architecture combines AI-assisted triage with other components commonly required in an internal IT support environment, including SLA monitoring, notications, ticket history, real-time communication, and operational analytics. Rather than treating AI as an independent chatbot, NIRNAYA places AI within the ticket-processing workow.

   The major contributions of the proposed system are:

   • An AI-assisted workow for internal IT support in which ticket triage is treated as a core system function.

   • A processing pipeline covering ticket classication, pri- ority prediction, department recommendation, agent rec- ommendation, summarization, and resolution assistance.

   • Integration of AI recommendations with SLA monitoring, notications, ticket history, and operational analytics.

   • A human-in-the-loop mechanism that allows support per- sonnel to review, accept, or modify AI-generated recom- mendations.

   • A clear distinction between an ITSM-focused platform such as NIRNAYA and general-purpose project and issue- management systems.

2. Problem Definition and Motivation

   A conventional IT support workow can be represented as

   T C P D A R, (1)

   where T represents the submitted ticket, C represents clas- sication, P represents priority selection, D represents de- partment selection, A represents assignment, and R represents resolution.

   In this workow, several stages depend directly on human interpretation. Although human judgment remains important, repetitive manual processing can become inefcient when the number of incoming requests increases.

   The proposed system introduces an AI-assisted version of this workow:

   T AI {C, P, D, A, S} H R, (2)

   where S represents resolution assistance and H represents human validation.

   The purpose of this approach is not to remove human involvement. Instead, the AI layer is intended to reduce repetitive interpretation and provide support personnel with useful information before they begin working on a ticket. The nal decision remains with the human agent.

   This approach is particularly relevant for requests that contain enough contextual information to identify their likely category, urgency, or routing requirements. At the same time, ambiguous requests can still be reviewed and corrected by support personnel.

   /li>

3. Related Work

   Existing research has investigated the use of articial intelli- gence and machine learning for automated support-ticket pro- cessing. The base paper, AI-based Classication of Customer Support Tickets, focuses on the application of AI techniques for classifying customer support requests [1].

   Automated prioritization and routing have also been inves- tigated as important components of IT support workows. Such approaches aim to reduce manual effort in determining the urgency of tickets and directing them to suitable support resources [2].

   Another area of research focuses on the classication of unstructured IT service-desk tickets. These approaches analyze

   free-form ticket descriptions and attempt to assign them to predened categories [3].

   Machine-learning-based help desk systems have been pro- posed to support ticket processing and service management activities [4]. Similarly, AI-based approaches have been in- vestigated for improving the efciency of ticket classication [5].

   Research surveys on ticket automation have examined the broader use of automation in ticket classication, routing, and management [6]. Customer-support ticket categorization and prioritization have also been considered as important tasks for improving support workows [7].

   AI-enhanced ticket management systems extend automated classication by integrating AI capabilities into the broader ticket-management process [8]. Machine-learning approaches have additionally been applied specically to IT support ticket classication [9].

   Based on these research directions, NIRNAYA combines ticket classication, priority prediction, department routing, agent recommendation, SLA monitoring, and human valida- tion within a unied internal IT support workow. Unlike approaches that focus on individual ticket-processing tasks, the proposed system integrates multiple decision-support functions within a single internal IT support workow while retaining human validation before nal ticket-handling decisions.

4. NIRNAYA SYSTEM OVERVIEW

   NIRNAYA is designed for two primary categories of users: employees and IT support personnel. Employees use the system to submit requests and monitor their progress. Admin- istrators and support agents use it to review incoming requests, examine AI-generated recommendations, manage tickets, com- municate with employees, monitor SLA conditions, and ana- lyze operational information.

   The proposed architecture is divided into ve logical layers:

   1. Presentation Layer: React-based interfaces through which employees and support personnel interact with the platform.

   2. Application Layer: Node.js/Next.js services responsible for authentication, ticket operations, business rules, and workow orchestration.

   3. AI/NLP Layer: AI-based services used to interpret ticket descriptions and generate recommendations.

   4. Data Layer: PostgreSQL used to store users, tickets, ticket history, SLA information, AI predictions, noti- cations, and agent activity.

   5. Real-Time Communication Layer: WebSocket-based communication for live status updates, notications, assignment changes, and SLA alerts.

5. AI-Driven Ticket Processing

   1. Ticket Understanding

      The processing pipeline begins when an employee submits a request in natural language. The application validates the request and forwards the relevant information to the AI layer.

      AI / NLP

PostgreSQL

Fig. 1. Logical architecture of the proposed NIRNAYA system.

The AI output is then converted into a structured representation that can be used by the remaining components of the system.

A conceptual representation of the proposed processing approach is

Y = f (X, H, O), (3)

where X represents the ticket text, H represents available history and contextual information, O represents organiza- tional constraints, and Y represents the structured output generated by the AI layer.

This formulation reects the fact that ticket interpretation may depend not only on the text of the current request but also on available contextual and organizational information.

1. Ticket Classication

   The classication component identies the operational area associated with a ticket. Example categories include network, hardware, software, access, security, database, and email- related issues.

   The purpose of this stage is to transform a free-form employee description into structured information that can be used by the subsequent routing and prioritization stages.

2. Priority Prediction

   NIRNAYA provides a recommended priority based on the information contained in the request and its available context. For example, a problem preventing an employee from ac- cessing an essential organizational resource may require more immediate attention than a minor conguration request.

   The predicted priority is treated as a recommendation rather than an automatic nal decision. This distinction allows the support agent to consider additional circumstances that may not have been available to the AI system.

3. Department and Agent Recommendation

   After interpreting the nature of a request, NIRNAYA recom- mends a suitable department and support agent. A conceptual suitability function is represented as

   S(a, t) = w1K(a, t)+w2W (a)+w3D(a, t)+w4E(a, t), (4)

   where K represents skill or category compatibility, W represents workload suitability, D represents department com- patibility, and E represents other relevant experience signals. The weights in this formulation are design parameters and are not claimed to be experimentally optimized in the current study. The formulation instead illustrates how multiple factors can be considered when generating an agent recommendation.

   WebSockets

Node.js / Next.js

React Presentation

1. Summary and Resolution Assistance

   In addition to classication and routing, the proposed system can generate a concise summary of the reported problem. It can also provide an initial troubleshooting direction based on the information contained in the ticket.

   For example, if an employee reports that a VPN stopped working after an operating-system update, the system may suggest checking the VPN certicate or restarting the VPN service.

   Such suggestions are intended to help the support agent begin the investigation. They are not intended to replace technical expertise or guarantee a successful resolution.

   Employee Ticket

Category Prediction

Priority Prediction

Department

Agent Recommendation

SLA + Notication

Human Validation

Resolution + Feedback

Fig. 2. AI-assisted ticket decision workow of the proposed system.

1. Human-in-the-Loop Decision Model

   An AI response generated by NIRNAYA may contain the predicted category, priority, department, recommended agent, condence value, summary, and resolution guidance.

   Rather than automatically applying these recommendations, the proposed system presents them to the support agent. The agent can accept the recommendations or modify them according to the actual situation.

   The prediction record can be represented as

   P = (t, y, q, ), (5)

   where t represents the ticket, y represents the AI recom- mendation, q represents its condence or score, and records the human decision when available.

   Maintaining both the AI recommendation and the nal human decision provides useful information for later analysis. It can also help identify situations in which the AI sstem requires improvement.

   The decision process can therefore be summarized as:

   AI recommendation human validation nal decision

   This design keeps human judgment at the center of the nal support decision while allowing AI to handle repetitive analytical tasks.

2. SLA Monitoring and Real-Time Communication

   Service Level Agreements (SLAs) are important in IT support because they dene expected response or resolution timelines. NIRNAYA associates SLA information with ticket processing and can alert support personnel when a ticket requires attention.

   This creates a connection between day-to-day ticket pro- cessing and longer-term service analysis. Support administra- tors can use these insights to understand recurring categories of requests and identify areas that may require additional attention.

   1. Comparison With General-Purpose Issue Management

      NIRNAYA is not intended to be positioned as a replacement for general-purpose issue-management platforms. Platforms such as Jira are designed to support a broad range of activities, including software development, project planning, backlog management, boards, and team workows.

      NIRNAYA instead focuses specically on internal IT service operations. The difference is primarily in workow emphasis rather than the presence or absence of individual features.

      TABLE II Conceptual Comparison

      If D represents the SLA deadline and t represents the

      current time, the remaining SLA time can be represented as

      Dimension General Work Man- agement

      NIRNAYA

      RSLA = max(0,D t). (6)

      This value can be used to identify tickets that are approach- ing their SLA deadline and support timely escalation.

      Real-time communication is implemented using WebSock- ets. This allows status changes, ticket assignments, comments, notications, and SLA-related alerts to be communicated without requiring users to repeatedly refresh the application.

3. Database and Analytics

PostgreSQL is used to maintain the structured operational information generated by the proposed system. The major entities considered in the design are shown in Table I.

TABLE I

Representative NIRNAYA Data Entities

Entity Purpose

User Employee, administrator, or support-agent infor- mation.

Ticket Description, category, priority, status, assign- ment, and timestamps.

AIPrediction AI outputs, condence scores, and recommen-

dation context.

Ticket History Status changes, assignments, comments, and

workow events.

SLA Record SLA target, elapsed or remaining time, and

Primary purpose Projects and work items Internal IT support Core object Issue/work item IT support ticket

AI role Assistive capability Core triage layer Classication Congurable Central AI function Priority Workow attribute AI-assisted prediction Department routing Conguration-dependent Core recommendation Agent recommenda- Not central Core recommendation tion

IT SLA Congurable Central workow Resolution assistance Possible through exten- Integrated design goal

sions

AI feedback Not central Explicit feedback concept

Software development Strong focus Not primary focus Sprint/backlog Core concepts Not intended focus

IT analytics Congurable Core dashboard objective

The comparison therefore reects a difference in intended use. General work-management systems organize and track work across many types of projects, whereas the proposed NIRNAYA system is designed around the specic decision- making requirements of internal IT support.

1. Illustrative Scenario

   Consider the following employee request:

   My VPN stopped working after the Windows up- date, and I cannot access company resources.

   NIRNAYA can transform the free-form request into struc- tured information that may resemble the following:

   • Category: Network

   • Priority: High

     Notication

     escalation state.

     Workow and real-time alerts.

   • Department: IT Support

     Agent Activity Assignment and support-operation information

     for analytics.

     The structured information stored by NIRNAYA can also support operational analytics. Examples include ticket volume, the number of open and resolved requests, category and priority distributions, SLA status, agent activity, and resolution trends.

   • Recommended Agent: Rahul Sharma

   • Condence: 94% (illustrative)

   • Summary: VPN connectivity failed after a Windows update.

   • Resolution assistance: Verify the VPN certicate and restart the VPN service.

   The values in this example are illustrative and are not presented as measured benchmark results. The purpose of

   the example is to demonstrate how an unstructured employee request can be converted into actionable information for review by a support agent.

2. Security and Privacy Considerations

   Internal IT support tickets can contain sensitive operational information, including employee identities, device details, ac- cess requests, and information about organizational systems. Access to such information should therefore be controlled according to user roles and responsibilities.

   A practical deployment of the proposed system should in- clude secure API communication, input validation, appropriate credential protection, audit logging, and controlled access to ticket information.

   The use of external AI services introduces an additional privacy consideration. Information sent to an external model provider should comply with organizational security policies and data-minimization requirements. For highly sensitive de- ployments, organization-controlled or locally hosted models could be considered to reduce dependence on external AI services.

3. Discussion

   The central idea behind the proposed system is to reduce the manual effort involved in the early stages of IT support. Instead of waiting for a support agent to manually interpret every incoming request, NIRNAYA provides an initial struc- tured assessment that can be reviewed before further action is taken.

   An important aspect of the proposed architecture is the integration of AI with the broader ticket lifecycle. Classi- cation and recommendation are connected with routing, SLA monitoring, communication, and analytics rather than being isolated as a separate conversational feature.

   The human-in-the-loop approach also provides a practical balance between automation and human judgment. Support agents remain responsible for the nal decision, while the AI system acts as a decision-support mechanism.

   However, the effectiveness of such a system depends on the quality of the input data, the AI model, organizational policies, and the quality of the available contextual information. These factors need to be considered when evaluating the proposed system in a real operational environment.

4. Limitations

   The proposed system has several limitations.

   First, AI recommendations may be inaccurate when a ticket is ambiguous, incomplete, or poorly described. Second, dependence on external AI services may introduce latency, cost, availability, and privacy concerns. Third, organizational categories, support procedures, and SLA policies can change over time and may require corresponding updates to the system.

   Another important limitation is the absence of a large- scale quantitative evaluation in the current study. Therefore, the proposed system does not claim specic improvements in

   response time, resolution time, classication accuracy, or SLA compliance.

   <>These measures require controlled experiments using rep- resentative IT support ticket datasets and clearly dened evaluation criteria.

5. Future Work

   Several areas can be explored in future versions of the proposed system.

   1. Empirical evaluation: Measure classication, priority prediction, department recommendation, and agent rec- ommendation using representative datasets.

   2. Human feedback integration: Analyze agent correc- tions to identify common AI errors and use the feedback to improve future recommendations.

   3. Workload-aware assignment: Incorporate current agent workload, availability, and skill proles into the assignment process.

   4. Knowledge-base integration: Connect the system with an approved organizational knowledge base to provide more context-aware troubleshooting assistance.

   5. Explainable recommendations: Provide explanations for why a particular category, priority, department, or agent was recommended.

   6. Privacy-preserving AI: Investigate locally hosted or organization-controlled models for sensitive IT support environments.

   7. Advanced SLA analytics: Extend the system to provide escalation prediction, SLA-risk analysis, and service- quality trends.

6. Conclusion

The proposed system, NIRNAYA, is an AI-driven internal IT Service Management platform designed to assist with the decision-intensive stages of employee IT support. It combines natural-language analysis with ticket classication, priority prediction, department and agent recommendation, summa- rization, resolution assistance, SLA monitoring, real-time com- munication, and operational analytics.

The primary contribution of the proposed system is the integration of these capabilities into a single support workow. AI is used as a decision-support component rather than as a replacement for support personnel. Human agents can review and modify recommendations before the nal action is taken. The proposed architecture is specically oriented toward internal IT operations, which distinguishes its workow from general-purpose project and issue-management platforms. Rather than attempting to replace such platforms, NIRNAYA focuses on the particular challenges of understanding, priori-

tizing, routing, and assisting with employee IT requests.

The proposed work establishes the architecture and work- ow for the NIRNAYA platform. Future evaluation using representative ticket datasets will be necessary to measure its classication performance, routing accuracy, operational efciency, and impact on SLA-related outcomes.

Acknowledgment

The authors would like to thank the Department of Com- puter Science and Engineering, Acharya Institute of Technol- ogy, for providing the academic environment, resources, and guidance required to carry out this work.

References

1. M. Truss and S. Bo¨hm, AI-based Classication of Customer Support Tickets: State of the Art and Implementation with AutoML, arXiv preprint arXiv:2406.01789, 2024, doi: 10.48550/arXiv.2406.01789.

2. M. Almarzooqi, Automated Prioritization and Routing of IT Support Tickets Using Machine Learning, M.S. thesis, Rochester Institute of Technology, RIT Dubai, Dubai, UAE, 2025.

3. S. P. Paramesh, C. Ramya, and K. S. Shreedhara, Classifying the Unstructured IT Service Desk Tickets Using Ensemble of Classiers, in Proc. 3rd Int. Conf. Computational Systems and Information Tech- nology for Sustainable Solutions (CSITSS), 2018, pp. 221227, doi: 10.1109/CSITSS.2018.8768734.

4. F. Al-Hawari and H. Barham, A machine learning based help desk sys- tem for IT service management, J. King Saud Univ. – Comput. Inf. Sci., vol. 33, no. 6, pp. 702718, 2021, doi: 10.1016/j.jksuci.2019.04.001.

5. S. Bonechi, G. Palma, M. Caronna, M. Ugolini, A. Massaro, and A. Rizzo, Enhancing Customer Support in Banking: Leveraging AI for Efcient Ticket Classication, Procedia Comput. Sci., vol. 246, pp. 128137, 2024, doi: 10.1016/j.procs.2024.09.235.

6. A. Zangari, M. Marcuzzo, M. Schiavinato, A. Gasparetto, and A. Albarelli, Ticket automation: An insight into current research with applications to multi-level classication scenarios, Expert Syst. Appl., vol. 225, Art. no. 119984, 2023, doi: 10.1016/j.eswa.2023.119984.

7. C. S. Kanimozhi Selvi, S. Jyothi Shri, R. Prashanthini, Neelamegan, and R. Sanjay, Customer Support Ticket Categorization and Priori- tization Using Natural Language Processing, in Proc. 3rd Int. Conf. Futuristic Technology (INCOFT), vol. 3, pp. 690698, 2025, doi: 10.5220/0013640400004664.

8. S. Jain, A. Gupta, and K. Neha, AI Enhanced Ticket Management System for optimized Support, in Proc. 4th Int. Conf. AI-ML Systems (AIMLSystems), 2024, pp. 17, doi: 10.1145/3703412.3703433.

9. L. S. Benitez Pereira, R. Pizzio, S. Bonho, L. M. F. de Souza, and A. C. Arnoni Junior, Machine Learning for Classication of IT Support Tick- ets, in Proc. Int. Conf. Cyber Management and Engineering (CyMaEn), 2023, doi: 10.1109/CyMaEn57228.2023.10051041.