Paper: arXiv 2506.18942

Authors: Simon Hatzesberger, Iris Nonneman

Abstract

This article explores the potential of generative AI (GenAI) to support actuarial practice through four implemented case studies. It situates these case studies within the broader evolution of artificial intelligence in actuarial science, from early neural networks and machine learning to modern transformer-based GenAI systems. The first case study illustrates how large language models (LLMs) can improve claim cost prediction by extracting informative features from unstructured text for use in the underlying supervised learning task. The second case study demonstrates the automation of market comparisons using Retrieval-Augmented Generation to identify, extract, and structure relevant information from insurers’ annual reports. The third case study highlights the capabilities of fine-tuned vision-enabled LLMs in classifying car damage types and extracting contextual information from images. The fourth case study presents a multi-agent system that autonomously migrates actuarial legacy code from R to Python and validates the translation against the original code’s outputs. In addition to these case studies, we outline further GenAI applications in the insurance industry. Finally, we discuss the regulatory, security, dual-use and fraud, reproducibility, privacy, governance, and organisational challenges associated with deploying GenAI in regulated insurance environments.

Complexity vs Empirical Score

  • Math Complexity: 2.5/10
  • Empirical Rigor: 6.5/10
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: This paper presents four practical GenAI case studies tailored to actuarial science, emphasizing implementation over mathematical theory. It excels in clarity and reproducibility, offering fully functional end-to-end Jupyter notebooks hosted via the International Actuarial Association, while providing a solid empirical baseline using synthetic insurance datasets.

Research Flowchart

  flowchart TD
    A[Research Goal: Advance Actuarial Practice via GenAI] --> B[Inputs: Unstructured Text, Reports, Images & Legacy Code]
    B --> C[Methodology: 4 Actuarial GenAI Case Studies]
    C --> D1[CS1: Text Feature Extraction for Claim Cost Prediction]
    C --> D2[CS2: RAG-based Automation of Market Comparisons]
    C --> D3[CS3: Fine-Tuned Vision-LLMs for Car Damage Classification]
    C --> D4[CS4: Multi-Agent System for R-to-Python Code Migration]
    D1 & D2 & D3 & D4 --> E[Outcomes: Enhanced Predictions, Automation & Regulatory Challenges Framework]

Cited References (66)

  1. ActuaryGPT : Applications of Large Language Models to Insurance and Actuarial Work (2024) — link
  2. On the Opportunities and Risks of Foundation Models (2022) — link
  3. The Actuary and IBNR (1972)
  4. AudioLM : A Language Modeling Approach to Audio Generation (2023)
  5. A Primer on Generative AI for Actuaries (2024) — link
  6. Teaching Large Language Models to Self-Debug (2024)
  7. AI – I s there a future for actuaries? (2024)
  8. Effective Statistical Learning Methods for Actuaries II: Tree-Based Methods and Extensions (2020)
and 58 more
  1. Statistical Learning Algorithms Applied to Automobile Insurance Ratemaking (2003)
  2. Agentic Retrieval-Augmented Generation: A Survey on Agentic RAG (2025) — link
  3. From Local to Global: A Graph RAG Approach to Query-Focused Summarization (2025) — link
  4. Artificial Intelligence Governance Principles: Towards Ethical and Trustworthy AI in the European Insurance Sector (2021) — link
  5. Artificial Intelligence Governance Framework (2025) — link
  6. The Impact of Artificial Intelligence along the Insurance Value Chain and on the Insurability of Risks (2022) — link
  7. Stochastic Claims Reserving in General Insurance (2002)
  8. Regulation (EU) 2024/1689 of the European Parliament and of the Council Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act) (2024) — link
  9. Boosting Insights in Insurance Tariff Plans with Tree - Based Machine Learning Methods (2021)
  10. MetaGPT : Meta Programming for a Multi-Agent Collaborative Framework (2024)
  11. Gen AI Risks for Businesses: Exploring the Role for Insurance (2025) — link
  12. SWE-bench : Can Language Models Resolve Real-World GitHub Issues ? (2024)
  13. Dual-use implications of AI text generation (2023) — link
  14. Classifying and Addressing the Diversity of Errors in Retrieval-Augmented Generation Systems (2026) — link
  15. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (2020)
  16. INS-MMBench: A Comprehensive Benchmark for Evaluating LVLMs’ Performance in Insurance (2024) — link
  17. A Platform for the Biomedical Application of Large Language Models (2025)
  18. Distribution-free Calculation of the Standard Error of Chain Ladder Reserve Estimates (1993)
  19. The World of Generative AI: Deepfakes and Large Language Models (2024) — link
  20. Inference for the Generalization Error (2003)
  21. G eneralized L inear M odels (1972)
  22. Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence (2023) — link
  23. The Risks of Using ChatGPT to Obtain Common Safety-Related Information and Advice (2023) — link
  24. OWASP Top 10 for Large Language Model Applications (2025) — link
  25. A Survey on LLM-based Multi-Agent Systems : Workflow , Infrastructure , and Challenges (2024) — link
  26. Automated Car Damage Assessment Using Computer Vision: Insurance Company Use Case (2024)
  27. Building AI Agents with LLMs , RAG , and Knowledge Graphs: A Practical Guide to Autonomous and Modern AI Agents (2025)
  28. Large Language Models for Code Generation : A Comprehensive Survey of Challenges , Techniques , Evaluation , and Applications (2025) — link
  29. AI in actuarial science – a review of recent advances – part 1 (2021)
  30. AI in actuarial science – a review of recent advances – part 2 (2021)
  31. An AI Vision for the Actuarial Profession (2024) — link
  32. The Credibility Transformer (2025) — link
  33. Learning representations by back-propagating errors (1986)
  34. Convolutional Neural Networks for vehicle damage detection (2022)
  35. Attention is All you Need (2017)
  36. AutoGen : Enabling Next-Gen LLM Applications via Multi-Agent Conversation (2024)
  37. From Generalized Linear Models to Neural Networks, and Back (2019) — link
  38. Data Analytics for Non-Life Insurance Pricing (2023) — link
  39. AI Tools for Actuaries (2025) — link
  40. Framework of BERT - Based NLP Models for Frequency and Severity in Insurance Claims (2023)
  41. Copyright in Generative Deep Learning (2022) — link
  42. Evaluating the Code Quality of AI-Assisted Code Generation Tools: An Empirical Study on GitHub Copilot, Amazon CodeWhisperer, and ChatGPT (2023) — link
  43. Actuarial Applications of Word Embedding Models (2020) — link
  44. BERT -based NLP Techniques for Classification and Severity Modeling in Basic Warranty Data Study (2022) — link
  45. A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions (2025) — link
  46. FinanceBench : A New Benchmark for Financial Question Answering (2023) — link
  47. Vehicle Damage Detection Using Artificial Intelligence: A Systematic Literature Review (2025) — link
  48. Large Language Model-Based Agents for Software Engineering: A Survey (2026) — link
  49. AI Revolution in Insurance: Bridging Research and Reality (2025) — link
  50. Artificial Intelligence Risk Management Framework (AI RMF 1.0) (2023) — link
  51. Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (2024) — link
  52. The Future of Credit Underwriting and Insurance Under the EU AI Act: Implications for Europe and Beyond (2025) — link
  53. Actuarial Professionalism Considerations for Generative AI: A Professionalism Discussion Paper (2024) — link
  54. Formalizing and Benchmarking Prompt Injection Attacks and Defenses (2024) — link
  55. SONAR 2025: New Emerging Risk Insights – How Deepfakes, Disinformation and AI Amplify Insurance Fraud (2025) — link
  56. Video Generation Models as World Simulators (2024) — link
  57. ImageNet Classification with Deep Convolutional Neural Networks (2012)
  58. GloVe : Global Vectors for Word Representation (2014)

0 of 66 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.