On Thursday, 24 September, the insurance and actuarial community will gather at CCT Venues Smithfield in London for the Institute and Faculty of Actuaries' AI and Emerging Technologies Symposium 2026. Dylogy is proud to take part: CEO and co-founder Aurélien Couloumy will present alongside Waswate Ayana of Convex, sharing new research on how agentic AI can turn casualty claims documents into structured, causal risk intelligence. The symposium brings together actuaries, data scientists, risk professionals, and business leaders for a full day exploring how AI is reshaping the profession, and it lines up closely with the problem Dylogy was built to solve.
Date: Thursday, 24 September 2026, 08:00–19:00 BST
Venue: CCT Venues Smithfield, Two East Poultry Ave, Smithfield, London, EC1A 9PT
Registration: Booking and pricing
From claims narratives to causal graphs
Aurélien and Waswate's session, "From Claims Narratives to Causal Graphs: An Agentic Pipeline for Casualty Actuarial Intelligence," tackles a problem that sits at the heart of Dylogy's work. Casualty claims, particularly in professional liability, encode complex chains of alleged failures, contributing factors, and cascading consequences, scattered across narrative reports, expert assessments, loss runs, legal filings, and photographs. Actuarial practice still tends to collapse this rich, multimodal information into flat tables, leaving one of insurance's most information-dense data sources largely unexploited.
The session presents a production-grade pipeline that combines LLM-based multimodal extraction, a formal ontology, and directed acyclic graph (DAG) construction within a multi-agent architecture. Vision-language models and large language models jointly extract causal entities and relationships across document formats, producing graphs that encode full liability chains. A domain-specific ontology defines actors, professional duties, alleged breaches, and damages, along with causal, temporal, and aggravating relationships, while an agentic orchestration layer coordinates data preparation, graph construction, and confidence validation, with human-in-the-loop checkpoints built in for actuarial governance.
The talk will walk through direct applications for professional liability: enriching loss development factor estimation by conditioning on graph-identified loss mechanisms, surfacing latent loss drivers across professions to support underwriting appetite calibration, and querying causal patterns for proactive risk prevention. Aurélien and Waswate will also address the practical hurdles of getting there, including context window constraints on large graph processing, the complexity of causal abstraction, cost-performance trade-offs, and the change management required to operationalise these techniques within actuarial teams.
What it means for underwriters, reinsurers, and risk teams
This is the same challenge Dylogy's platform addresses every day: insurance documents hold more signal than tabular data alone can capture, and unlocking it requires AI built specifically for the structure and stakes of insurance work. For underwriters, reinsurers, brokers, risk managers, and legal teams navigating growing document volumes, the symposium's focus on agentic AI, governance, and practical deployment offers a preview of where the profession is heading, and where the tools to get there are already being built.
Join us in London
Whether you're just starting to explore AI or already piloting agentic pipelines, the AI and Emerging Technologies Symposium 2026 offers a full day of practical insight for actuaries, risk professionals, and insurance leaders alike. Register through the IFoA's booking page to secure your seat, and if you'll be in London on 24 September, get in touch with the Dylogy team. We'd welcome the chance to compare notes on where agentic AI is heading in insurance.
| Domaine d'Application | Bénéfices Clés grâce aux Graphes de Connaissances |
| Prévention | Identifier les causes racines communes à plusieurs sinistres pour proposer des actions préventives. Détecter les causes secondaires pour limiter la propagation des dommages. |
| Tarification | Proposer des segmentations plus fines des assurés en fonction des chaînes de causes réelles. Mieux évaluer l'impact d'une exclusion de garantie (par exemple, "Exclusion de l’usure de machines"). |
| Provisionnement | Comprendre les dynamiques d'évolution des pertes et identifier les causes qui, étonnamment, ne mènent à aucun coût, pour affiner les réserves financières. |
