Ex-Head of AI at Qover, where he built the 100% autonomous claims programme powering Revolut at greater-than-human accuracy. 50,000+ AI claims solved (and counting). Now leading product and engineering at Ellis.
We're building the autonomous claims layer European insurance has been waiting for.
The industry's automation effort has stalled in endless POCs. Generic AI doesn't understand the terrain. Consulting practices don't ship decision engines. Ellis exists to close that gap — and to do it with the trust insurance demands.
Operators who've shipped autonomous claims at scale before.
Ex-Head of Insurance Partnerships at Qover, where he launched the partnerships function from scratch and scaled it from $0 to $8m in revenue, dealmaking with large insurers, brokers and MGAs. Now leading commercial at Ellis.
Operators, AI engineers, claims experts.
A small, in-person team based in Paris. We're hiring deliberately — depth over headcount.
Four hard problems have kept end-to-end claims automation out of reach. Ellis is built to close them.
The problem in AI claims handling today
Today's use cases cover the easy 20% of the workload.
AI use cases tackle intake, triage, and minor verifications. The heavy lifting on complex coverage decisions remains untouched.
ROI is marginal until handlers can delegate the time-consuming work.
T&Cs and PIGs are the visible tip of the ruleset.
Each insurer interprets the same check differently. The rest lives nowhere on paper — only in handlers' heads as implicit, tribal knowledge.
Without this knowledge, accuracy caps at 80%, while the bar sits at 99%.
A claim is not a document — it's a stack of tools.
Claim history, fraud scans, depreciation rules — each in a different system. If the AI can't access them, the assessment is half-finished.
Handlers won't trust an analysis they have to re-stitch by hand.
Insurers can't deploy a claims AI without regulatory sign-off.
Compliance and security gate every deployment — rightfully so. Regulators demand specific documentation, and the bar is high.
The most accurate system in the world doesn't go live without it.
How Ellis addresses it
End-to-end handling, at 99% accuracy.
50+ specialised agents run every check — evidence, coverage, fraud, depreciation, outcome — from first notice to closed decision. Not the easy 20%. The whole claim.
Encodes written and unwritten rules.
A dedicated calibration layer ingests handler feedback and edge-case arbitrages, updating decision rules without retraining any model. Tribal knowledge becomes the engine's operating logic — versioned and reversible.
Compliant out of the box.
EU-processed, ISO 27001-aligned, GDPR and EU AI Act ready. DPA, sub-processor register, and decision explainability documentation available from day one.
Let's have a conversation.
Whether you'd like to see Ellis in action, have questions about fit or compliance, or just want to exchange ideas — we're happy to talk. Drop us a line at contact@ellisclaims.ai and you'll hear back from one of the founders directly.
Ellis SAS · Registered in the European Union.
Let's talk.
30-minute call with the founders. We'll do a live walkthrough on a line of business of your choice.
Talk to the founders