AI-Driven Student Support System using Retrieval-Augmented Generation in Academic ERP

Abstract

The fast pace digital evolution in the field of higher education has made the use of Enterprise Resource Planning (ERP) and Learning Management Systems (LMS) inevitable for all sorts of learning and administrative functions. Unfortunately, the currently available systems do not offer a consolidated solution as they require interacting with numerous applications in order to manage attendance, assignments, schedule, and other academic operations. In addition to being inefficient, such a design creates challenges in terms of data synchronization and visibility as well as increasing the burden on administration. In this paper, Campus Nova, a web platform that allows inte-grating academic ERP functionality and an AI-driven chatbot, is proposed. Built using React.js and Firebase, the system employs serverless computing to facilitate data synchronization and scaling. The proposed web application provides a dual-access interface with separate modules for tracking attendance, managing assignments, scheduling classes, and tracking students’ academic performance. One of the significant contributions lies in using a Retrieval-Augmented Generation (RAG) enabled chatbot that allows interacting with the data through a natural language processor. Furthermore, the system uses the concept of differential attendance updates with the help of session locking mechanism.

Introduction

The paper presents Campus Nova, an intelligent campus management system that combines traditional Enterprise Resource Planning (ERP) functions with a Retrieval-Augmented Generation (RAG)-based conversational AI assistant.

Background and Motivation

Modern educational institutions widely use ERP and LMS platforms to manage enrollments, attendance, schedules, assignments, and other academic processes. While these systems have improved data integration and administration, they still rely on conventional dashboards that require users to manually navigate multiple modules. Existing systems also lack intelligent, real-time guidance and conversational interaction.

Recent advances in Artificial Intelligence (AI) and Large Language Models (LLMs) enable natural language interactions, but standalone LLMs often suffer from hallucinations and lack access to up-to-date institutional data. To address these issues, RAG combines LLMs with external data retrieval mechanisms, improving accuracy and relevance.

Proposed Solution: Campus Nova

Campus Nova integrates ERP functionality with a RAG-powered chatbot capable of accessing real-time academic data. The system uses:

• A React-based frontend for user interaction.
• Firebase Firestore as a cloud-based database.
• A RAG pipeline that retrieves relevant academic information and provides context-aware responses through an LLM.

Key Contributions

1. Development of a unified academic ERP platform for students and faculty.
2. Integration of a RAG-based conversational assistant for real-time academic query handling.
3. Implementation of attendance management with differential updates and session control.
4. Use of a scalable serverless architecture based on modern web technologies.

Literature Review Findings

The review highlights:

• ERP systems improve educational administration and analytics.
• Educational chatbots enhance accessibility but often rely on static knowledge bases.
• RAG systems improve response accuracy by retrieving relevant external information before generating answers.
• Existing research rarely combines ERP systems, RAG technology, and multi-role academic workflows into a single platform.

Research Gap

Current ERP systems lack advanced conversational capabilities, while existing educational chatbots typically cannot access real-time academic data. There is limited research on integrating:

• ERP data management,
• RAG-based contextual response generation,
• Student and instructor workflows within one system.

System Architecture

Campus Nova follows a three-tier architecture:

1. User Interface Layer – React-based dashboard and chatbot interface.
2. Data Storage Layer – Firebase Firestore database storing users, courses, attendance, assignments, and timetables.
3. AI Reasoning Layer – RAG-enabled chatbot that retrieves relevant academic information and generates responses using an LLM.

Methodology

The query-processing workflow involves:

1. Receiving a natural language query from the user.
2. Detecting the user's intent (e.g., attendance, assignments, timetable).
3. Retrieving relevant academic records from Firestore.
4. Creating a structured context from the retrieved data.

Conclusion

This study introduced a platform known as Campus Nova, which is a smart campus management tool that unites all the functionality of an academic ERP system and a retrieval-augmented generation conversational assistant. This approach eliminates problems associated with traditional academic plat-forms through the creation of non-fragmented user interfaces that allow users to easily interact with their academic accounts. The experiments have shown that the developed system makes working with academic information easier by making the data accessible, requiring less effort from the user. The combination of real-time retrieval systems with large language models proved to be very efficient. Among other future work that will be done includes use of AI algorithms to predict attendance, creation of a mobile app, ability to push alerts, and development of a data analysis tool for better insight. Integration of a multi-agent intelligent agent system can add more intelligence to the platform.

References

[1] Brown and R. Patel, “Enterprise resource planning systems in higher education: Architecture and challenges,” IEEE Transactions on Education, vol. 67, no. 2, pp. 145–154, 2024. [2] T. Pasaribu and A. Suryanto, “Digital transformation of university erp systems: A cloud-based approach,” IEEE Access, vol. 12, pp. 52 340–52 352, 2024. [3] H. Daryanto, “Smart academic management systems for modern univer-sities,” IEEE Transactions on Learning Technologies, vol. 18, no. 1, pp. 55–66, 2025. [4] Y. Qin and X. Zhang, “Retrieval-augmented generation for knowledge-intensive nlp tasks,” IEEE Transactions on Knowledge and Data Engi-neering, vol. 35, no. 10, pp. 10 120–10 134, 2023. [5] J. Wuttke and M. Braun, “Enhancing conversational ai with retrieval-augmented generation,” IEEE Transactions on Artificial Intelligence, vol. 6, no. 3, pp. 899–910, 2025. [6] H. Kim and J. Park, “Deep learning based academic performance pre-diction system,” IEEE Transactions on Learning Technologies, vol. 18, no. 2, pp. 145–156, 2025. [7] A. Alahmari and S. Khan, “Artificial intelligence framework for smart campus systems,” IEEE Access, vol. 12, pp. 87 541–87 555, 2024. [8] J. Lee and S. Kim, “Educational chatbots in digital learning environ-ments,” IEEE Transactions on Learning Technologies, vol. 16, no. 3, [9] pp. 332–343, 2023. [10] S. Maity and D. Roy, “Intelligent academic chatbot using natural language processing,” IEEE Access, vol. 13, pp. 20 125–20 137, 2025. [11] S. Uppalapati and R. Kumar, “Transformer-based conversational agents for academic assistance,” IEEE Access, vol. 13, pp. 60 215–60 229, 2025. [12] A. Pathak and R. Pandey, “Conversational ai for academic management systems,” IEEE Transactions on Education, vol. 68, no. 1, pp. 44–52, 2025. [13] L. Xinyue and T. Zhang, “Retrieval-augmented generation for knowledge-intensive ai systems,” IEEE Transactions on Knowledge and Data Engineering, vol. 36, no. 6, pp. 2205–2218, 2024. [14] S. Rao and P. Sharma, “Hybrid retrieval-augmented ai systems for enterprise knowledge platforms,” IEEE Transactions on Artificial In-telligence, vol. 6, no. 4, pp. 1120–1132, 2025. [15] A. Zhalgasbayev and B. Nurkhanov, “Vector-based retrieval methods for generative ai systems,” IEEE Access, vol. 12, pp. 99 214–99 227, 2024. [16] R. Selvam and V. Krishnan, “Ai-based intelligent digital assistants for organizational systems,” IEEE Access, vol. 13, pp. 41 025–41 039, 2025.

Copyright

Copyright © 2026 Aditya Tiwari, Amritanshu Ranjan, Sandeep Gupta. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.