|
Organised by the EAA – European Actuarial Academy in cooperation with the Lietuvos Aktuarų Draugija
Generative artificial intelligence (GenAI) is rapidly changing how actuarial work can be done: from turning unstructured information into model-ready data, to accelerating analysis, documentation, and communication. Large language models (LLMs) enable workflows that go far beyond “chatting with ChatGPT” – but using them responsibly in actuarial practice requires a clear understanding of capabilities, limitations, evaluation, and robust integration into existing processes.
This two-day, hands-on seminar focuses on advanced, practically implementable applications of GenAI in actuarial science. After establishing a practical foundation (how modern LLMs work, where they succeed and fail, and how to assess output quality), participants will work through a series of case studies that reflect typical insurance realities: messy data, document-heavy processes, and the need for auditability, traceability, and human oversight. Throughout the programme, we will connect concepts such as prompting patterns, structured outputs, function calling, retrieval-augmented generation (RAG), fine-tuning, multimodal capabilities, and agentic AI to concrete actuarial use cases.
The core of the seminar is built around five applied case studies, each combining a clear business goal with hands-on implementation:
- Claims cost prediction enriched by text: using LLMs to derive meaningful features from unstructured claim descriptions, and integrating these into predictive models to improve performance and insight.
- Automated market comparisons with RAG: building an LLM-supported workflow that searches and synthesizes information from annual reports, product documents, policy wordings, etc. to speed up structured comparisons.
- Multimodal claims support for motor insurance: leveraging (fine-tuned) vision-enabled LLMs to classify car damage types from images and extract relevant contextual information for downstream processes.
- Agentic data analysis and reporting: demonstrating how an LLM-based multi-agent system can autonomously explore a dataset, run analyses, and draft a coherent report of key findings – with human control points and quality checks built in.
- Report generation and quality assurance with GenAI: designing a report pipeline that drafts actuarial narratives from inputs (results, tables, assumptions) and applies built-in checks (consistency, completeness, citation/trace-back to sources, and red-flag detection) before human review.
We conclude with an outlook and discussion that presents additional practical applications beyond the main case studies, without going into full implementation detail. We also address the challenges of applying generative AI in insurance and discuss future developments and their implications for actuarial work.
The seminar will be held in person, giving participants the opportunity to learn on site alongside other actuarial professionals, exchange ideas directly, and receive immediate support from the lecturers. The venue is the 4-star hotel Courtyard Vilnius City Centre (details below). The evening event on the first day is giving participants the opportunity to connect, discuss practical questions, and build their professional network in an informal setting.
Early-bird discount is available for bookings made by 1 August 2026.
|
