Dylogy launches Claims Discover, an agentic risk learning system built on(re)insurers' claims portfolios

PRESS RELEASE — Lyon, France — June 10, 2026

By applying agentic AI to transform thousands of unstructured claims files into actionable risk signals, Claims Discover enables (re)insurers and brokers to understand why losses occur and to make sharper claims management and underwriting decisions across their portfolio.

Claims data: an underexploited asset

Claims teams process hundreds of thousands of files each year: expert reports, legal correspondence, photographs, loss notifications. Yet the overwhelming majority of these documents remain unexploited beyond the settlement of the claim itself.

Current claims management systems tell teams what happened: who, where, when, which line of business. They say nothing about the origin of the loss, the aggravating factors that drove up costs, the recurring sources of coverage disputes, or shifts in the underlying risk profile.

The gap between the information that actually sits within claims documents and what existing tools are able to extract from it carries a real cost: underwriting guidelines that remain incomplete, approximate actuarial assumptions, under-efficient claims handling decisions, and prevention policies that are permanently lagging behind loss experience.

Claims Discover adds an intelligence layer on top of existing systems

Claims Discover is not a claims management system. It connects to existing tools and adds a dedicated agentic analytics layer, drawing on a business taxonomy and on loss event chains reconstructed file by file. It processes an entire claims portfolio regardless of format, language or complexity, extracting structured, reusable knowledge that builds over time.

The solution operates at four levels:

• ‍Building a business taxonomy: working alongside subject matter experts, Claims Discover generates a set of descriptive attributes tailored to each line of business.‍
• Reconstructing loss causation and chronology: for each claim, the solution rebuilds the graph linking the initial cause, consequences, contributing factors, mitigation levers and aggravating conditions, drawing on all documents in the file. Every element is traceable toits source document, by design reducing the risk of hallucination.‍
• Portfolio-level aggregation: results are normalised against the organisation's own referential and standards, and consolidated to surface underlying trends across the book.‍
• Detection of emerging loss patterns and recommendations: the most frequent or most severe causal sequences are identified and fed into concrete recommendation outputs for each function: claims handler, underwriter, actuary, loss prevention specialist, and others.

Each of these functions derives direct operational benefit.

• The causes and context of a loss finally become legible: the claims handler can more accurately match them against contractual cover, enabling controlled indemnity payments and full client satisfaction.
• The detection of emerging loss events and their associated aggravating factors translates into actionable underwriting recommendations: from the risk questionnaire to the renewal offer, and for the development of new products.
• The data collected feeds aggregation methods and supports a deeper understanding of how losses develop over time: sharpening actuarial assumptions for reserving, technical pricing and fraud modelling.
• The causes of losses and their associated remediation or mitigation levers continuously inform prevention policies, and can even become a genuinely differentiating service offering for insurers' direct clients.

"For years, risk knowledge has been accumulating in claims files without ever being structured. By combining a business ontology with causal graphs built at portfolio scale, Claims Discover turns that knowledge into an exploitable asset for underwriters, actuaries and claims teams alike. This is no longer reporting. It is risk learning."— Aurélien Couloumy, CEO and co-founder of Dylogy

Measured results across live portfolios

Available in production since January 2026, Claims Discover has already demonstrated its impact across live portfolios. Observed results to date include:

• 95% reduction in time spent on manual claims coding and classification;
• Over 200 analytical attributes extracted per claim file;
• 1,500 claims files processed and structured in a single day;
• Around fifty recurring loss patterns identified per portfolio on average each month;
• Measurable impact on the adjustment of non-life reserve development factors;
• Dozens of actionable recommendations generated each week across claims, underwriting and actuarial teams.

The solution has been deployed across approximately twelve lines of business, in both France and the UK, covering property and casualty (home, motor), specialty lines(energy, marine, aviation) and life & health (health, income protection).

In early deployments, teams were able to surface recurring trends previously invisible in their data, opening the way to targeted revisions of policy wording and underwriting criteria.

Claims Discover sits within Dylogy's broader product ecosystem, which also includes contract analysis and decision-support solutions already in use by some ten(re)insurers and brokers. The solution is available today.

About Dylogy

Founded in 2023, Dylogy accelerates decision-making for insurers, reinsurers and brokers through agentic AI technologies that structure and leverage the value of their documentary knowledge. Its SaaS solution addresses key challenges in accelerating contract analysis, enhancing claims management reliability and strategic risk monitoring through dedicated business applications and an integrated ERM platform. Labelled French Tech ScaleUp Excellence, the French start-up counts a dozen clients (including Convex, Howden, Groupama and Gedeon), achieved nearly €1M in revenue in 2025 (50% international) and raised €2M in pre-seed funding.