Meet your AI Agent for Loss Run Processing.

Built for loss-run math

Loss runs aren't documents — they're tables of money. InsurGPT™ reads them like an underwriter: roll up incurred when components are listed separately, normalize across carrier formats, and surface the numbers that actually drive the quote.

Insurance Math Built In | Calculated

Calculates total incurred when paid, reserves, and expenses are listed separately. Rolls up frequency and severity by policy period and coverage line so the underwriter sees the shape of the loss history, not just the raw rows.

Confidence Scoring on Every Field | Scored

Every extracted value carries a calibrated confidence score. Low-confidence items route to human review automatically — no silent guesses make it to the underwriter or into the rating engine.

X-Ray Mode Traceability | Traceable

Every extracted number is linked back to its exact location in the source document. Reviewers click a value and jump to the cell, page, and coordinates it came from — verification in seconds, not minutes.

Trained on insurance, not the internet

InsurGPT™ isn't a general model with an insurance prompt on top. It's an ensemble of specialized models trained on 300M+ non-public insurance documents — loss runs, ACORDs, SOVs, schedules — across dozens of carrier-specific formats.

Domain-Native Corpus | Insurance-only

300M+ non-public insurance documents in the training set — loss runs, ACORD forms, SOVs, schedules of values, and carrier exports — not scraped web text dressed up for the industry.

Carrier-Format Fluency | Multi-format

Reads loss runs from dozens of carrier templates without per-carrier configuration. Field names, column orders, and footnote conventions are recognized automatically across the long tail of formats.

Ensemble of Specialists | Multi-model

Dozens of specialized models, each tuned for a specific document type and extraction task. The right model reads the right artifact — not one generalist guessing at all of them.

Built for the underwriting desk

Extraction is the start, not the finish. The output lands inside the underwriter's existing workflow — structured, reviewable, and ready for the rating engine — with the audit trail intact.

Structured Output | Quote-ready

Delivers loss-run data as clean structured records mapped to your schema. Drops straight into the rating engine, the submission system, or the underwriter's worksheet without a re-keying pass.

Human-in-the-Loop Review | Reviewable

Low-confidence fields and edge-case formats route to a reviewer queue with the source document pre-anchored to the cell in question. Reviewers correct, approve, and move on.

Full Audit Trail | Auditable

Every value, every confidence score, every reviewer action is logged with a timestamp and a link to the source artifact. Compliance and reinsurance audits run against a complete provenance chain.