GapFollow-form drift at Layer 3L3 · L2
Arbor adds a PFAS exclusion the umbrella lacks, opening a $5M–$15M hole for PFAS claims. ARB 21 07 · p.2NSU 00 01 §IV · p.6
The whole tower, read as one program.
Shared-and-layered programs read layer by layer — concurrency, attachment points, follow-form, the wording deltas between markets — every finding cited to the layer. On your book, and on the accounts you’re pursuing.
Tower review — four layers, reconciled against the lead form and the schedule
Representative screen · specimen programReportsTower reviewHartline — 2026 casualty program
Export
Draft memo
Hartline — 2026 casualty program
Hartline Manufacturing · $25M excess of primary · 4/1/2026 – 4/1/2027
4 layers7 policies5 carriers3 issues
Program towerLimits excess of primary
$25M
L4$10M xs $15MQuota shareFollow form
Ironbridge Excess50% · $5M p/o $10M4/15/26 – 4/15/27
Summit National50% · $5M p/o $10M4/15/26 – 4/15/27
$15M
L3$10M xs $5MFollow-form drift
Arbor CasualtyFollow form + 1 exclusion
ARB 21 07PFAS exclusionnot in NSU 00 01
$5M
L2$5M umbrellaxs primaryLead form
Northstar Specialty$25K SIR
$0
L1PrimaryKeystone Mutual4/1/26 – 4/1/27
CGL$2M / $4MCG 00 01
Auto$1M CSLCA 00 01
EL$1M / $1M / $1MWC 00 00 00
Concurrency findings5 checks across 4 layersAll layers
GapPolicy period misaligned at Layer 4L4 · L3
Ironbridge and Summit incept 4/15/26; $10M xs $15M is bare from 4/1 to 4/15. Ironbridge Dec · p.1Arbor Dec · p.1
UpdateSchedule of underlying: expired autoL2 · L1
Schedule A lists auto BA 25-10448, expired 4/1/26; the renewal is BA 26-11873. NSU Sched. A · p.4Keystone BA Dec · p.1
ConfirmSIR not carried into excess drop-downL3 · L2
Northstar retains $25K where no underlying applies; Arbor's drop‑down is silent on it. NSU Decl. Item 4 · p.1ARB XS §III.B · p.5
ConsistentQuota share: both follow formL4
Ironbridge 50% and Summit 50% follow the Northstar form; shares total 100%. IBX 00 01 §I · p.2SNX 10 §I · p.2
Programs fail between the layers.
Dates that don’t line up, follow-form that quietly doesn’t follow, five words that change how a claim pays. Coverage-law commentators note that excess carriers invoke date discrepancies to deny claims — and the manual defense is a specialist’s week, rationed to the loudest accounts.
Reconciles every layer against the lead and the schedule.
Concurrency, attachment points, sublimits, definitions, notice provisions — each layer read by the specialist for its line, then cross-read as one program, including the cross-policy gaps no single-document read can see.
Reads the wording deltas and says what they do to a claim.
Manuscript against specimen, market against market — with functional equivalence respected: different words that do the same job are matched, and familiar terms a form quietly redefined are not.
Drafts the recommendation — and the teardown.
The program analysis in plain client language: what is inconsistent, what it means at claim time, what to ask the markets for. Pointed at a pursuit, it is the incumbent-program teardown — cited, in your brand.
The job, defined
“For our complex risk, they are built in towers… multiple policies by various carriers layered on top of each other, and they just follow the language below. The entirety of the coverage is defined by the complete stack.”
Customer, recorded call
With Qumis
“This particular wording comparison has come up two or three times — a game changer. It’s something I’m always asking for and we’ve never had.”
Placement specialist, recorded call
Asked for, unprompted, before it had a name.
191professionals at 34 organizations asked Qumis program-structure questions — concurrency, attachment points, layering, follow-form, quota share — in the last 180 days, with no named program report shipped.
Qumis production data, 180 days to 26 August 2026.
The entirety of the coverage is defined by the complete stack.
Teams that read towers here usually run these next.
The questions buyers ask first.
We already run general AI assistants. Why use Qumis on towers?
Use both — our largest customers do. A general assistant can describe what a pollution exclusion does; it has no method for a tower: testing each policy’s retroactive date independently, auditing the umbrella’s schedule of underlying, catching follow-form drift, or walking adjacent policies through the cross-policy patterns where one form’s exclusion pushes a claim into another’s.
Our specialty desk already does this.
On the accounts it can reach. Qumis runs the same read on every layered program in the book — and on the incumbent programs your producers are pursuing — so the specialist’s week goes to judgment, not to the first pass.
We don’t say “program design” — we just need tower review.
Then this is your use case. Tower review, wording comparison, and concurrency are the work; “program structure and design” is what it adds up to when the recommendation is drafted.
How big a program can it handle?
Towers of any practical depth. Documents are read whole rather than in retrieval snippets, production programs run past two thousand pages in a single document, and package policies are decomposed into their coverage parts and read per line.
Underneath the six is Coverage Intelligence — the system of judgment Qumis is building for commercial insurance, where every reviewed decision makes the next one better.