Intelligent Support Optimization Framework for Virtual Banking Assistants using ML on AWS Cloud

Abstract

Virtual Banking Assistants (VBAs) are integral to modern customer support; however, they frequently suffer from \"contextual amnesia\" and an inability to resolve high-complexity queries, leading to service latency and user frustration. This study introduces an Intelligent Support Optimization Framework aimed at seamlessly integrating automated responses with human intervention. The framework employs a staged Machine Learning pipeline for enhanced decision-making: TF-IDF vectorization paired with Multinomial Logistic Regression provides rapid intent classification, while Decision Trees and XGBoost are integrated to evaluate conversation failure conditions and risk-based urgency. To ensure seamless continuity, the architecture leverages Amazon DynamoDB for state persistence and AWS Lambda for serverless orchestration, allowing the system to preserve customer context during escalation. Implementation on AWS cloud infrastructure ensures high scalability and low-latency execution. Performance evaluations using Precision, Recall, and F1-score confirm that the suggested framework significantly minimizes query redundancy while optimizing the routing process to live representatives The proposed hybrid ML approach effectively enhances the operational efficiency and reliability of virtual banking systems in a production environment

Introduction

Virtual Banking Assistants (VBAs) have become an essential part of digital banking by providing automated, real-time customer support for services such as account management, payment tracking, and issue resolution. However, existing VBA systems face several challenges, including loss of conversation context during session interruptions or transfers to human agents, rigid session timeouts, inaccurate query routing, and the inability to identify urgent issues such as fraud or suspicious transactions. These limitations often result in repeated explanations, delayed resolutions, and reduced customer satisfaction.

To address these issues, this study proposes an Intelligent Support Optimization Framework that combines machine learning with Amazon Web Services (AWS) cloud infrastructure. Unlike conventional chatbots that focus only on intent recognition, the proposed system emphasizes context preservation, intelligent query routing, real-time risk assessment, and seamless interaction management. The framework uses TF-IDF feature extraction and Multinomial Logistic Regression for accurate intent classification, Decision Tree algorithms for routing queries to the appropriate banking department, and XGBoost to assess query urgency and automatically escalate high-priority cases such as fraud or unauthorized transactions. AWS Lambda enables serverless processing, while DynamoDB stores conversation history to maintain continuity even if sessions are interrupted or transferred between agents.

The literature review shows the evolution of banking chatbots from rule-based systems to machine learning and NLP-based models. Although methods such as Logistic Regression, Naïve Bayes, SVM, Random Forest, and XGBoost have improved intent classification, most existing systems still lack effective context management, dynamic risk assessment, and flexible session handling. Similarly, while AWS-based cloud architectures improve scalability and performance, they do not fully address conversational continuity.

The proposed framework follows a structured workflow: users submit banking queries through a chatbot interface, the text is preprocessed and converted into numerical features using TF-IDF, intents are classified using Logistic Regression, queries are routed through a Decision Tree model, and XGBoost evaluates their priority for automatic escalation when necessary. Conversation history is continuously stored in the cloud, allowing customers to resume interactions without repeating previous information.

Conclusion

This research presented an Intelligent Support Optimization Framework for Virtual Banking Assistants using Machine Learning and AWS cloud technologies to improve customer support efficiency in digital banking environments. The proposed framework was designed to address major operational challenges present in existing banking support systems, including increased customer waiting time, repeated query explanations during agent transfers, incorrect department routing, and conversational context loss caused by session interruptions. The framework integrated TF-IDF feature extraction with Multinomial Logistic Regression for accurate intent classification, Decision Tree-based intelligent routing for efficient department allocation, and XGBoost for urgency and risk assessment of banking queries. AWS Lambda was utilized for scalable serverless orchestration, while Amazon DynamoDB enabled conversational context preservation and session continuity during customer interactions.Experimental evaluation demonstrated that the proposed framework successfully improved banking customer support performance by reducing response delays, minimizing incorrect query routing, and eliminating repeated customer explanations. The DynamoDB based context preservation mechanism ensured seamless continuation of conversations even when support agents changed or sessions disconnected, thereby significantly improving customer experience. The framework achieved strong classification with an accuracy of 94.2%, precision of 93.5%, recall of 92.8%, and an F1-score of 93.1%, the results confirm that merging machine learning models with cloud-based conversation systems effectively enhances intelligent banking support. The proposed architecture introduces a scalable, Intelligent, and customer-centric solution for next-generation digital banking assistance systems. The integration of Machine Learning and AWS cloud services demonstrates strong potential for improving real-time banking customer support operations, conversational continuity, and automated service management in modern financial institutions. Future enhancements may include multilingual conversational support, voice-enabled banking assistants, sentiment analysis, real-time fraud analytics, and advanced deep learning-based conversational intelligence for further improvement of intelligent banking support systems.

References

[1] Panel Agnieszka Nowak-Brzezinska a,Dawid Drabek bA \"hybrid knowledge-based machine learning approach for ticket classification in the AWS cloud\" in scienceDirect Volume 270,2025 . [2] R. M. Rachani and R. Ravish, \"Cloud Based IDS-Anomaly Detection Using Machine Learning,\" 2025 1st International Conference on AIML-Applications for Engineering & Technology (ICAET), Pune, India, 2025, pp.1-6,doi:10.1109/ICAET63349.2025.10932232 [3] E. Gunawan, A. J. Amadeo, C. Kamardi and D. Suhartono, \"Virtual Assistant System in Banking Application With Fine-Tuned LLM and RAG Approach,\" 2025 International Conference on Informatics, Multimedia, Cyber and Information System (ICIMCIS), Jakarta, Indonesia, 2025, pp. 961-966, doi: 10.1109/ICIMCIS68501.2025.11327124 [4] K. Marappan, M. N, P. S, V. R, S. R and C. K, \"Exploring Machine Learning Benefits for Healthcare with AWS Lambda Technology,\" 2025 11th International Conference on Communication and Signal Processing (ICCSP), Melmaruvathur, India, 2025, pp. 1831-1836, doi: 10.1109/ICCSP64183.2025.11088750 [5] G. Solaimalai, J. M. Shanthi, S. K. S, P. Khabiya, K. Geetha and G. Talele, \"Natural Language Processing in Virtual Assistants Current Approaches and Challenges,\" 2024 International Conference on Science Technology Engineering and Management (ICSTEM), Coimbatore, India, 2024, pp. 1-5, doi: 10.1109/ICSTEM61137.2024.10560644 [6] V. Gadipally, P. Kunekar, M. Goel, A. Patil, D. Abdurakhimova and S. Kakad, \"Natural Language Processing for Automated Customer Service in Banking,\" 2025 World Skills Conference on Universal Data Analytics and Sciences (WorldSUAS),Indore,India ,2025,pp.1-6 doi: 10.1109/WorldSUAS66815.2025.11199260. [7] A. Szappan, R. Fleiner and E. Kail, \"AI-Based Complaint Handling Challenges in the Financial Sector,\" 2025 IEEE 25th International Symposium on Computational Intelligence and Informatics (CINTI), Budapest, Hungary, 2025, pp. 000707-000712, doi: 10.1109/ CINTI67731.2025.11311812. [8] Mangipudi, Phanidra. (2025). \"Ai Powered Chatbots Vs Human Agents, A Compatible Study On Customer Satisfaction. International Journal Of Computer Engineering & Technology\". 16. 11-36. 10.34218/Ijcet_16_02_002. [9] Vijaya Kittu Manda; Khaliq Lubza Nihar, \"Artificial Intelligence Applications in the Indian Financial Ecosystem,\" in Artificial Intelligence for Risk Mitigation in the Financial Industry , Wiley, 2024, pp.133-157, doi: 10.1002/9781394175574.ch.6. [10] L. L. Méndez-Gutiérrez, J. E. Espinoza-Castro and E. J. Jiménez-Pachas, \"Artificial Intelligence and Customer Experience: A Systematic Review Across Industries 2021 - 2025,\" 2025 IEEE International Conference on Technology Management, Operations and Decisions (ICTMOD), Glasgow, United Kingdom, 2025, pp. 1-6, doi: 10.1109/ICTMOD66732.2025.11372032. [11] A. Tiwari, R. K. Pandey, S. Mishra, A. Singh, S. Singh and V. K. Chaudhary, \"Machine Learning Pipelines with AWS SageMaker for Customer Churn Analysis,\" 2025 International Conference on Sustainability, Innovation & Technology (ICSIT), Nagpur, India, 2025, pp. 1-6, doi: 10.1109/ICSIT65336.2025.11294323.. [12] Eustaquio-Jiménez, Roberto & Durand-Azurza, Mercedes & Gamboa Cruzado, Javier & León Morales, María & Ruiz, Nancy & Montoya, Reyna & Céliz, N.. (2024).\"Chatbots for customer service in financial entities\"—A comprehensive systematic review. Journal of Infrastructure, Policy and Development. 8. 10122. 10.24294/jipd10122. [13] R. Agrawal, S. Desai, D. Dholwani, N. Kedari and A. Banerjee, \"Artificial Intelligence/Machine Learning Driven Decision making in Business Analytics for Financial Sector using Ensemble Machine Learning Techniques,\" 2024 IEEE 3rd World Conference on Applied Intelligence and Computing (AIC), Gwalior, India, 2024, pp. 371-376, doi: 10.1109/AIC61668.2024.10731028. [14] S. Mishra and A. Tiwari, \"Medical Insurance Cost Prediction using AWS Sage-Maker,\" 2024 International Conference on Augmented Reality, Intelligent Systems, and Industrial Automation (ARIIA), Manipal, India, 2024, pp. 1-6, doi: 10.1109/ARIIA63345. [15] M. K. Ravindranathan, D. Sendil Vadivu and N. Rajagopalan, \"Cloud-Driven Machine Learning with AWS: A Comprehensive Review of Services,\" 2024 International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE), Bangalore, India, 2024, pp. 1-8, doi: 10.1109/ IITCEE59897.2024.10468032. [16] Asad Abdo, Shafiz Mohd Yusof\" Exploring the impacts of using the artificial intelligence voice-enabled chatbots on customers interactions in the United Arab Emirates\"Vol 12, No 4,Dec.2023.DOI: http://doi.org/10.11591/ijai.v12.i4. [17] A. Ali, A. Irfan, A. Raza and Z. Memon, \"Banking in the Digital Age: Predicting Eligible Customers through Machine Learning and AWS,\" 2024 IEEE 1st Karachi Section Humanitarian Technology Conference (KHI-HTC), Tandojam, Pakistan, 2024, pp. 1-6, doi: 10.1109/ KHI-HTC60760.2024.10482026. [18] Nikita and G. Parashar, \"Brain Stroke Detection and Prediction Using Machine Learning Approach: A Cloud Deployment Perspective,\" 2023 International Conference on Circuit Power and Computing Technologies (ICCPCT), Kollam, India, 2023, pp. 1705-1714, doi: 10.1109/ICCPCT58313.2023.10245699. [19] B. Prabadevi, R. Shalini, B.R. Kavitha,\"Customer churning analysis using machine learning algorithms\",International Journal of Intelligent Networks,Volume 4,2023,Pages 145-154,ISSN 2666-6030,https://doi.org/10.1016/ j.ijin.2023.05.005. 2024.11051598. [20] A. Kiran,I.J.Kumar, P.Vijayakarthik, S.K.L.Naik and T.Vinod,\"Intelligent Chat Bots: An AI Based Chat Bot For Better Banking Applications,\" 2023 International Conference on Computer Communication and Informatics(ICCCI), Coimbatore, India, 2023, pp. 1-4 doi: 10.1109/ICCCI56745.2023.10128582.

Copyright

Copyright © 2026 Jogulapati Bhagya Suma, Dr. Dasari Haritha. 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.