Rating and filings

How the price of an admitted risk actually moves — and why the house’s medical- professional-liability play runs through an underwriting bulletin rather than a rate filing. This page is the mechanism underneath the healthcare wedge.

Filed rating plans: the slow machinery

An admitted carrier’s prices live in a filed rating plan: base rates, classification tables, and rating factors filed with each state’s regulator and supported by actuarial justification (Vol 1: rating & pricing). Changing the plan — a new credit, a new class, a new factor — means a filing: actuarial exhibits, regulator review, possible objections, one state at a time. For a countrywide commercial line that is a multi-year grind, and it is why “we’ll get carriers to file an attestation credit” is a strategy that dies of old age. STABLE

Schedule rating: the discretion already filed

Most commercial filed plans carry a schedule-rating section: the underwriter may debit or credit an individual risk within a bounded range, for named characteristics — management quality, loss-control programs, record-keeping, safety practices. The bounds are filed; the exercise of the discretion is not. A schedule mod is applied at the desk, account by account, with documentation — no regulator in the loop. STABLE

VERIFY — confirm before you rely on it

the schedule-mod discretion commonly cited is roughly ±25%, but the filed range varies by state and by line — confirm the specific carrier’s filed plan before building a pitch on it · source: the carrier’s state rating-plan filings (state DOI / SERFF).

The instrument that turns discretion into practice is an underwriting bulletin — internal guidance from underwriting management to desk underwriters: a risk presenting attested AI telemetry under an open standard may warrant a schedule credit under loss control / management. A bulletin issues in days, not years. Nothing is filed, because nothing new is being used — the discretion was filed all along. GROUNDED · Glacis

Bulletins, endorsements, filings — the speed table

Instrument	What changes	Who signs off	Clock
Underwriting bulletin	How already-filed discretion (schedule mods) is exercised	Underwriting management	Days–weeks
Endorsement / guideline change	Terms on a given account — e.g., an exclusion carve-back, using paper from the carrier’s endorsement library	Underwriting management	Weeks
Rate filing	The plan itself — new credits, classes, factors	State regulator(s), with actuarial support	Quarters–years, per state

The ordering of the clocks is the strategy. STABLE

Exclusion carve-backs

When a market gets nervous it writes exclusions; when it learns to see the risk, it sells the cover back. A carve-back is an endorsement that returns a defined slice of excluded coverage, on conditions the underwriter can verify. The decision to offer one is an underwriting-guideline change — the fast lane again, not a filing campaign. STABLE

The live application: medical professional liability. Carriers of the Aesculap Mutual class hold ambient-scribe exposure today — their insured physicians already dictate into ambient AI documentation tools (Dictaphage Health and its competitors) — with zero visibility into how those systems behave. Meanwhile AI exclusions are propagating through the standard forms, opening a gap between what physicians do and what their policies say. GROUNDED · Glacis

VERIFY — confirm before you rely on it

the pace and scope of AI exclusions entering standard professional-liability forms (reportedly widening through 2026), and the commonly cited ~$12B annual premium of the US medical professional liability line · source: ISO form circulars via carrier filings; NAIC annual-statement line aggregates.

One credit — attested ambient AI documentation → schedule credit, exclusion waived — converts an existing multi-billion-dollar line into the house’s distribution: once one respected carrier gives the credit, every scribe vendor must buy attestation to keep its customers insurable, and every competing carrier must explain why it cannot see what the first one sees. GROUNDED · Glacis

The sequenced play

1. Now — the bulletin conversation. Schedule credit plus carve-back, opened immediately: it is the longest cycle in the book and the biggest prize, so it starts first.
2. Later — the filed credit, brought to the regulator with case-study actuarials, at the point where formalizing is defensive for the carrier: its competitors are quoting against the credit, and the filing locks the pricing vernacular in on the carrier’s terms.

The order is the point. The fast paths generate the loss experience that makes the slow path filable; run in reverse, the sequence is a multi-year wait for a credit no actuary can yet support. GROUNDED · Glacis

Questionnaire underwriting — and why it fails for AI

The incumbent method for underwriting AI risk is the proposal-form questionnaire: ask the applicant to describe its own model governance, testing, and controls, then rate off the answers. Self-reported, point-in-time, unverifiable — asking drivers whether they are safe drivers. GROUNDED · Glacis

For AI it fails on two axes at once:

• Time. The answers describe the system on the day the form was signed; a model update the following week silently invalidates them — Vol 2’s opening argument, the assessment that “expires the moment a model updates in production” (the unpriced risk).
• Trust. Every answer is a management assertion — the bottom rung of the denominator trust ladder. Nothing is independently attested.

Chancel Specialty launched an affirmative AI liability cover underwritten exactly this way. The house’s read is unsentimental: Chancel is now holding AI severity it cannot observe, and its early questionnaire-underwritten loss experience becomes the case for attestation either way — if the book runs clean, the credit conversation is cheap; if it runs hot, “we are the telemetry between you and your first bad loss year” writes itself. That pitch — embedding verification inside the cover a carrier has already launched — is page 02’s opening move. GROUNDED · Glacis

Next

→ 02 Sharing the upside

Drill this page →24 bank questions stand behind what you just read. Check it while it’s warm.