AI for insurance teams, from claim intake automation to underwriting risk scoring.

Yes, with conditions. AI used for risk assessment and pricing in life and health insurance is high-risk under Annex III §5(b). That triggers obligations on risk management, data governance, technical documentation, record-keeping, transparency, human oversight, accuracy, robustness, and cybersecurity. We build conformity-assessment scaffolding into the system from week one: an ISO 42001-aligned governance memo, the technical documentation pack the regulation requires, an append-only audit log, and a documented human-in-the-loop step for any output that affects an underwriting decision. Non-life lines may sit outside Annex III §5(b) but EIOPA principles and GDPR Article 22 still apply.

Decisions producing legal or similarly significant effects (a quote refusal, a claim denial, a premium loading above a threshold) cannot be made solely on automated processing without explicit safeguards. We architect the system so any decision in scope of Article 22 is presented to a human reviewer with the model output, the feature-level explanation, and a one-click path to override. The override and the reviewer ID are written to the audit log. This satisfies the supervisory expectation and gives the policyholder a meaningful appeal path.

A first engagement targets one workflow with a measurable baseline, runs 4 to 12 weeks to production, and lands as a single signed-off system inside one core platform or claims surface. Common scopes are: claim-intake automation across one or two product lines; underwriting risk scoring on one book of business; fraud-triage scoring on a defined claim category; policyholder service automation on a defined query taxonomy. Submit a project with the workflow you have in mind and the rough volume, and we scope and price the discovery phase before any code is written.

Yes. The delivery layer is built around your core platform, not the other way around. We ship integrations with Guidewire ClaimCenter, PolicyCenter and BillingCenter, Duck Creek OnDemand, Sapiens IDIT and CoreSuite, Tia, and the major broker platforms. For systems without a modern API we build a queue-based bridge with idempotent writes and a manual reconciliation interface. The audit log writes regardless of where the data lands, so you can prove lineage even when the downstream core cannot.

Fairness is treated as a measurable property of the model, tested against the protected and proxy variables that matter in your jurisdiction. We define a fairness metric set during discovery (typically a mix of statistical parity, equal opportunity and disparate impact thresholds), test the model against it pre-production and on every retrain, and ship the report to your model-risk and compliance functions. EIOPA expects this posture as a matter of supervisory practice, and the AI Act codifies it. We do not ship insurance models that have not been fairness-tested.

By default, all processing and storage runs in EU regions on infrastructure under EU jurisdiction. We support regional pinning when a regulator or contract requires it (Germany-only, France-only, Lithuania-only, US-only). Original policy and claim documents land in immutable EU object storage with hashes recorded in the audit log. We do not train any model on your data, full stop. If your contract requires US-resident processing for a US-only book, we expose that as an explicit configuration toggle, never a default.

Production-grade deployments see field-level error rates of 0.4 to 1% on routine FNOL documents after the first three weeks of evaluation tuning, against a 2 to 4% human-only baseline reported in industry studies. Accuracy depends on document quality, line of business, and the breadth of the evaluation set. We do not claim a single accuracy number across all carriers and lines. We baseline first, target a specific delta against your current process, and report against it weekly through the pilot.

Pricing is set after the discovery sprint, against your specific workflow, book of business and core-platform integration surface. We do not publish a flat rate because the scope variation across insurance AI is wide: a single-product FNOL extraction system on a uniform corpus is a different build from a cross-jurisdiction underwriting risk-scoring stack with explainability for supervisors. Submit a project with the workflow and rough volume, and we come back with a discovery proposal within one business day.