Smart Fraud Analytics in Modern InsurTech Using a Hybrid Generative and Agentic AI Architecture
DOI:
https://doi.org/10.5281/zenodo.20423200Keywords:
Generative Artificial Intelligence,Agentic AI Systems,Fraud Detection,InsurTech,Machine Learning,Deep Learning,Explainable AI (XAI),Anomaly Detection,Predictive Analytics,Natural Language Processing (NLP),Multi-Agent Systems,Risk Assessment,Synthetic Data Generation,Real-Time Fraud Detection,Intelligent Decision Systems.Abstract
In a hybrid generative and agentic artificial intelligence architecture, generative AI generates fraud-detection signals based on historical events while agentic AI provides decision-suggestion commands to the human stakeholder in-the-loop. This synergistic approach overcomes the limitations of pure generative and agentic frameworks in detecting fraud patterns in modern InsurTech platforms. Rapidly growing digitalization and increased dependence on digital channels increase the susceptibility of insurance platforms to behavioral fraud. Advanced technologies such as AI and Machine Learning (ML) have the potential to mitigate the risk of billions of dollars lost through these frauds. To harness the power of both generative and agentic AIs, a novel hybrid system is proposed. Key stakeholders of the system include the InsurTech venture for which the system is being built, customers of the insurance platform, the development team, and the operational team responsible for maintaining and managing the AI pipelines.
Data aspects of the architecture encompass novel feature-engineering techniques, a sophisticated data-pipeline approach, and an incremental learning mechanism to prevent performance deterioration due to concept drift. An unsupervised-ML-based segmentation model has been instantiated for the generation of training datasets and enables scalable continuous-signal generation in a real-time production environment. The quality of signals in the fraud-detection domain is not always suitable for direct operationalization. Hence, the operational design leverages the hybrid nature of the architecture by enabling human evaluation of the suggestion signals in a real-time mode. The primary fraud-detection application of the system is claims related, although it has the capability to generate fraud signals for policy underwriting, channel fraud, and behavioral fraud use cases as well.
References
[1]Adelugba, E. (2026). Artificial intelligence-driven fraud detection models in healthcare insurance claims processing. International Journal of Engineering Technology Research & Management, 10(3), 1–12.
[2]Deepa, V., & Pushpa, B. (2026). AI-based fraud detection in insurance claim in India. International Journal of Advanced Multidisciplinary Research and Educational Development, 2(1), 1260–1265.
[3]Yang, H., Liu, Y., & Chen, X. (2026). A review of artificial intelligence for financial fraud detection. Applied Sciences, 16(4), 1931.
[4]Zafar, U., Khan, M., & Lee, J. (2026). Methodological challenges in explainable AI for fraud detection. Artificial Intelligence Review, 59(2), 1–28.
[5]Lakshmanan, R. (2026). Harnessing artificial intelligence to improve fraud detection in insurance claims. IJETRM, 10(3), 45–52.
[6]Lakshmanan, R. (2026). Natural language processing for anomaly detection in insurance documentation. IJETRM, 10(3), 53–60.
[7]Odufisan, O. I., Abhulimen, O. V., & Ogunti, E. O. (2025). Harnessing artificial intelligence and machine learning for fraud detection and prevention. Journal of Economics and Technology, 12(1), 100127.
[8]Jain, A., Kulkarni, R., & Lin, S. (2025). Explainable AI in big data fraud detection. arXiv preprint arXiv:2512.16037.
[9]Almalki, F., & Masud, M. (2025). Financial fraud detection using explainable AI and stacking ensemble methods. IEEE Access, 13, 1–15.
[10]Zhang, C. J., Gill, A. Q., Liu, B., & Anwar, M. J. (2025). AI-based identity fraud detection: A systematic review. Computers & Security, 135, 103497.
[12]Ke, Z., Zhou, S., Zhou, Y., Chang, C. H., & Zhang, R. (2025). Detection of AI deepfake fraud using GAN-based models. IEEE Transactions on Information Forensics and Security, 20, 1–12.
[13]Sheth, A., Roy, K., & Gaur, M. (2023). Neurosymbolic artificial intelligence: Why, what, and how. IEEE Intelligent Systems, 38(3), 56–62.
[14]Holzinger, A., Langs, G., Denk, H., Zatloukal, K., & Müller, H. (2019). Causability and explainability of AI. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 9(4), e1312.
[15]Hamet, P., & Tremblay, J. (2017). Artificial intelligence in medicine. Metabolism, 69, S36–S40.
[16]Holmes, J., Sacchi, L., & Bellazzi, R. (2004). Artificial intelligence in medicine. Annals of the Royal College of Surgeons of England, 86(5), 334–338.
[17]Ngai, E. W. T., Hu, Y., Wong, Y. H., Chen, Y., & Sun, X. (2011). Application of data mining techniques in fraud detection. Decision Support Systems, 50(3), 559–569.
[18]Baesens, B., Van Vlasselaer, V., & Verbeke, W. (2015). Fraud analytics using descriptive, predictive, and social network techniques. Wiley.
[19]Phua, C., Lee, V., Smith, K., & Gayler, R. (2010). A comprehensive survey of data mining-based fraud detection. arXiv preprint arXiv:1009.6119.
[20]Abdallah, A., Maarof, M. A., & Zainal, A. (2016). Fraud detection system: A survey. Journal of Network and Computer Applications, 68, 90–113.
[21]Kou, Y., Lu, C. T., Sirwongwattana, S., & Huang, Y. P. (2004). Survey of fraud detection techniques. IEEE International Conference on Networking, Sensing and Control, 749–754.
[22]Bolton, R. J., & Hand, D. J. (2002). Statistical fraud detection. Statistical Science, 17(3), 235–255.
[23]Carcillo, F., Le Borgne, Y. A., Caelen, O., & Bontempi, G. (2021). Streaming active learning strategies for real-life credit card fraud detection. Data Mining and Knowledge Discovery, 35, 1–25.
[24]Roy, S., & Sun, J. (2021). Deep learning detecting fraud in financial transactions. Expert Systems with Applications, 164, 113848.
[25]Douzas, G., & Bacao, F. (2022). Improving imbalanced learning through synthetic data. Information Sciences, 580, 104–119.
[26]Fiore, U., De Santis, A., Perla, F., Zanetti, P., & Palmieri, F. (2019). Using generative adversarial networks for fraud detection. Information Sciences, 479, 448–455.
[27]Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT Press.
[28]Brown, T. B., et al. (2020). Language models are few-shot learners. NeurIPS, 33, 1877–1901.
[29]Vaswani, A., et al. (2017). Attention is all you need. NeurIPS, 30, 5998–6008.
[30]Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers. NAACL-HLT, 4171–4186.
[31]Radford, A., et al. (2021). Learning transferable visual models from natural language supervision. ICML, 139, 8748–8763.
[32]Schick, T., et al. (2023). Toolformer: Language models can teach themselves to use tools. NeurIPS, 36.
[33]Minsky, M. (1986). The society of mind. Simon & Schuster.
[34]Russell, S., & Norvig, P. (2021). Artificial intelligence: A modern approach (4th ed.). Pearson.
[35]Sutton, R. S., & Barto, A. G. (2018). Reinforcement learning: An introduction (2nd ed.). MIT Press.
[36]Wooldridge, M. (2020). An introduction to multiagent systems (2nd ed.). Wiley.
[37]Bubeck, S., et al. (2023). Sparks of artificial general intelligence. arXiv preprint arXiv:2303.12712.
[38]OpenAI. (2023). GPT-4 technical report. arXiv preprint arXiv:2303.08774.
[39]Deloitte. (2025). AI-powered fraud detection in insurance. Deloitte Insights.
[40]European Insurance and Occupational Pensions Authority (EIOPA). (2025). AI governance and fraud detection guidelines in insurance. EIOPA Report.
Additional Files
Published
Issue
Section
License
This work is licensed under a Creative Commons
Attribution 4.0 International License (CC BY 4.0).
You are free to:
- Share: copy and redistribute the material
- Adapt: remix, transform, and build upon the material
for any purpose, even commercially.
Under the following terms:
Attribution — You must give appropriate credit to
the original author(s) and source.
Full license text: https://creativecommons.org/licenses/by/4.0/
