Inside an excess casualty submission: what InsightXtract extracts — and why it matters

Excess and umbrella casualty is one of the hardest lines to intake well. A single submission for a mid-market trucking or manufacturing account can run to a dozen files: the broker’s email with the ask, an ACORD 125, a schedule of underlying policies, a five-year exposure workbook, a per-vehicle fleet schedule, a statement of values, five years of loss runs from three carriers, and a set of financials. The information that actually decides the price — the trend in revenue and fleet size, the development on open claims, the attachment adequacy of the underlying tower — is buried across those files, in different formats, and it is almost never on any one page.

Most intake tools read one document at a time and hand back a flat list of fields. That is not how an excess account is underwritten. InsightXtract runs an agentic, multi-document pipeline that classifies each file, extracts to a per-document schema, consolidates everything into one unified account record, links it into an entity graph, and validates it against your rules — so the underwriter opens a single, coded, cited record instead of a stack of PDFs.

One submission, many documents, one connected record — classify, extract, consolidate, link, validate.

The idea that governs everything: the time series is the price

Before the categories, the single most important concept. In excess casualty, a snapshot is nearly useless — the trend is the risk. An account doing $88M in revenue tells you little; an account that grew from $46M to $88M in three years while its fleet went from 70 to 120 power units is a very different exposure, and its excess rate should reflect that. The same is true of losses: a $312K total incurred means one thing if it’s flat and another if it’s developing upward with open reserves.

So InsightXtract captures the exposure and loss data not as single values but as time series — the current term plus three to five prior years. In the tables below, the Captured column flags this explicitly:

• Current — captured once for this term.
• Current + prior — captured per year so the model sees the trend.
• Per-row — a table: one vehicle, location, underlying policy, or claim per row.
• Point-in-time — a single static fact.

A & B · Insured, account, and broker

Who the account is and who’s placing it — identity, legal structure, operations classification, and the distribution chain. Sources: ACORD, broker email.

Parameter	Captured	Why it matters
Named insured, DBAs, FEIN, entity type	Point-in-time	The contract party and unique account key — drives clearance, conflicts, and de-dupe.
NAICS · SIC · ISO GL class · operations description	Point-in-time	Class-based rating and appetite/knockout — the hazard grade beyond the code.
Years in business · parent / subsidiaries	Point-in-time	Stability credit and aggregation of exposure across a group.
Broker / agency · contact · wholesaler chain	Point-in-time	Distribution routing, binding authority, and correspondence.

C & D · Submission and requested coverage

The ask itself — term, transaction, and the exact shape of the excess/umbrella cover being requested. Sources: broker email, ACORD.

Parameter	Captured	Why it matters
Product · transaction (new/renewal) · effective & expiration	Current	Coverage form, workflow, term, and binding deadline.
Target / expiring premium	Current + prior	Price-to-beat and the rate-change baseline for a renewal.
Each-occurrence & aggregate limit	Current	The capacity being offered — the core of the ask.
Attachment point / SIR · layer position	Current	Buffer vs. high-excess — drives burn-through risk and price.
Follow-form vs. standalone · drop-down · sublimits · defense inside/outside	Per-row	Scope alignment with the underlying and true limit erosion.
Occurrence vs. claims-made · retro date · territory	Current	Trigger, tail exposure, and geographic terms.

E · The underlying tower — one row per policy

An umbrella is only as sound as what it sits over. InsightXtract reads the schedule of underlying policies as a table — one row each. Source: underlying-policy schedule, ACORD.

Parameter	Captured	Why it matters
Line (GL / Auto / Employers Liab / Foreign) · carrier · AM Best · admitted?	Per-row	What the umbrella follows and the security of the paper below it.
Policy number · effective / expiration	Per-row	Concurrency with the excess term — gaps are a knockout.
Underlying limits (per-occ / agg / CSL) · retention / deductible	Per-row	Attachment adequacy — is the umbrella actually sitting where the ask says?
Underlying premium · aggregate erosion	Per-row	Rate adequacy below, and impaired aggregates that pull the excess down.

F · Financial exposure — the exposure base, year by year

This is where the trend lives. Each metric is captured for the current term and three-to-five prior years, because the year-over-year change drives the excess rate far more than any single value. Sources: exposure workbook, financials, ACORD.

Parameter	Captured	Why it matters
Annual revenue / sales / gross receipts	Current + 5 yr	Primary exposure base — growth means rising exposure the rate must track.
Total payroll · employee count (FT/PT/seasonal)	Current + 5 yr	Injury and employers-liability exposure base and workforce-mix trend.
Subcontractor / 1099 spend	Current + prior	Contingent and transferred liability the umbrella may absorb.
Units produced / sold · square footage · admissions	Current + prior	The class-appropriate exposure unit for rating.
Foreign / international sales %	Current + prior	Foreign-liability and venue exposure that changes the tower.
Projected next-term exposure	Current	The forward-looking basis the layer is actually rated on.

G · Auto & fleet exposure — fleet summary + per-vehicle schedule

Auto is the most common driver of umbrella losses and the source of nuclear verdicts. InsightXtract captures the fleet summary (with prior-year counts for trend) and a row per vehicle from the vehicle schedule / ACORD 129. Sources: vehicle schedule, exposure workbook, ACORD.

Parameter	Captured	Why it matters
Total power units & trailers	Current + prior	Fleet size and growth — a direct frequency-trend signal.
Units by type (PPT / light / medium / heavy / tractor) & GVW class	Per-row	Severity mix — heavy trucks drive the biggest verdicts.
Radius of operation · annual mileage (total & per unit)	Current + prior	True road exposure and interstate-venue risk; miles-based rating.
DOT # · MC # · hazmat / placarded · CSA scores	Point-in-time	FMCSA identity, hazard escalation, and a regulatory safety signal.
Driver count · MVR standards · avg experience	Current	Driver quality — the number-one auto loss driver.
Per vehicle: year · make · model · VIN · GVW · cost new/ACV · class · use · radius · garaging	Per-row	The covered-unit schedule — per-unit rating and physical-damage value.

H, I & J · Premises, products, and geography

The rest of the exposure surface — where the operations physically sit, what they make and do, and which venues they touch. Sources: statement of values, ACORD, exposure workbook.

Parameter	Captured	Why it matters
Per location (SOV): address · occupancy · sq ft · owned/leased · construction · TIV	Per-row	Premises-liability footprint and slip-and-fall exposure.
Products / completed-ops description & aggregate · recall history	Current	Products-liability and recall severity the umbrella covers.
Subcontractor use · additional-insured status · contractual liability assumed	Current	Where liability actually lands — risk-transfer quality.
Per state: payroll · revenue · units · miles by jurisdiction	Per-row	Venue-weighted exposure and social-inflation / nuclear-verdict risk.

K · Loss history — five-year loss runs, per-claim + rollups

The single biggest pricing input. InsightXtract reads each claim as a row across the five-year loss run — often from several carriers — and computes the rollups that matter. Source: loss runs.

Parameter	Captured	Why it matters
Per claim: claim # · date of loss · line · state · cause · claimant · litigation	Per-row	Each loss’s identity, venue, nature, and escalation potential.
Per claim: incurred · paid · outstanding reserve · status (open/closed)	Per-row	Cost, development, and open exposure still on the books.
Total incurred & claim count by year · loss ratio · frequency · avg severity · largest loss	Current + 5 yr	Frequency and severity trend — rate adequacy and volatility.
Losses in the excess layer · open reserves · loss-run valuation date	Current	Burn-through relevance to the attachment and data freshness.

L, M & N · Prior insurance, risk quality, and financials

The context that separates two accounts with identical exposures — continuity, safety culture, and solvency. Sources: ACORD, broker email, financials.

Parameter	Captured	Why it matters
Expiring carrier · limits · retention · premium · prior excess carriers (5 yr) · continuity date	Current + prior	Rate-change baseline, coverage continuity, and claims-made tail.
Coverage gaps · declinations · cancellations / non-renewals	Prior 5 yr	Adverse-selection and moral-hazard flags.
Fleet safety / telematics · driver hiring standards · WC experience mod (EMR)	Current + prior	Safety culture — the proxy for future frequency.
Total assets · liabilities · net worth · revenue · EBITDA · D&B · bankruptcy history	Current + prior	Ability to fund the retention and collectibility of the SIR.

From fourteen categories to one connected record

Pulling these parameters out of eight documents is only half the job. The value is in consolidation: the insured named on the ACORD, the broker on the email, the underlying limits on the schedule, the fleet on the vehicle workbook, and the claims on three loss runs all describe one account. InsightXtract merges them into a single record by a declared source-of-truth priority (the ACORD wins over the email on insured details; the underlying schedule wins on attachment), and links the result into an entity graph — insured at the centre, connected to broker, submission, coverage, underlying policies, fleet, locations, and loss history.

Why it matters to the business

Comprehensive, structured, trend-aware extraction isn’t a data-entry nicety — it changes the economics and quality of the book:

• Better-priced risk. Capturing exposures and losses as a five-year time series — not a snapshot — lets you rate to the trend, which is where excess money is won or lost.
• Fewer missed exposures. The parameters that cause surprises — an impaired underlying aggregate, a hazmat fleet, a developing open claim, a coverage gap — are exactly the ones buried deep in the schedules. Extracting them by default means they reach the underwriter instead of the claim file.
• Faster quotes, more capacity. An underwriter opening a consolidated, cited record instead of a dozen PDFs triages and prices in minutes, not hours — more submissions handled without more headcount.
• Consistency and auditability. The same categories, coded the same way, every time — with provenance to the source document. That is the difference between a repeatable book and one that depends on which underwriter opened the file.
• Portfolio intelligence. Once every submission is coded to the same schema, the exposure and loss trends roll up — you can see fleet growth, venue concentration, and loss development across the whole book, not one account at a time.

The Excess Casualty agent extracts all of this today — the account, submission, coverage, and underlying tower; the per-state/class exposure schedule; the fleet broken down by vehicle type; the multi-year exposure and auto-unit history; and the loss run — consolidated into one coded record of 50+ fields and six linked tables, every value cited to its source document. The point this post makes is why it matters: the trend and the per-vehicle detail are exactly what move an excess price, so they’re captured by default. And because it’s all configuration — fields and tables in the agent’s output contract, not code — the schema keeps pace with what underwriters ask for.