Use caseCommercial property

Supporting documents for underwriting

The broker sends nine attachments and wants a quote in ten days. Two years of loss runs are missing, and nobody notices until the due date.

The pack is checked against your guidelines the hour it arrivesThe broker gets a precise request the next dayThe underwriter prices, and the manager approves above the limit

Actions on this submission
The assistantGuidelines check

The activity is within appetite and the values are under $50M. But the loss runs cover 3 years of 5, and the survey dates from March 2023.

source · 9 documents in the submission
The verdict on the pack

A piece is missing · the AI names it, with its source

01

What gets in the way

The short version

Forty to sixty property submissions reach a commercial lines team every week. Each one is an email with seven to fifteen attachments: an application, a schedule of values, loss runs, a survey report, certificates. Ridgeline Castings sent nine, and two years of loss runs were not among them. The underwriter learns that the day she finally opens the pack.

Within the hour it lands, the pack is read. The AI types each attachment, pulls the values and the losses into the submission, and checks the whole pack against your underwriting guidelines. The underwriter starts from the gaps and the figures, and keeps the price and the terms.

Loss runs 2023-2025.pdfViewer
Three years where five are due. The loss runs open beside the checklist of what the broker sent, and what he did not.
My submissionsCase management
Megan's week, by due date. Twelve submissions, the AI verdict as one more column, Ridgeline the largest.

An underwriter knows all four of these screens. Their order decides the outcome: the pack is read the hour it arrives, the broker is asked for the missing pieces while there is still time, and the quote is built on a complete file. The three acts below stay with Ridgeline Castings until the quote leaves, two days early.

02

The pack arrives

Megan Hayes has not opened the email yet. The pack is already filed, read and held against the guidelines.

Step 1Intake

The broker's email becomes a submission case

On Monday at 11:20, Mark Hollis of Hollis Risk Partners sends a renewal for Ridgeline Castings, a metal parts maker with two sites near Cleveland, Ohio. The email reaches the brokers' mailbox with nine attachments. Capture creates a submission case, links it to the broker and gives each attachment its type.

ApplicationSchedule of valuesLoss runsSurvey reportSprinkler certificate
Submission SUB-26-1187 · DocumentsCase management
Each attachment is typed on arrival, so the checks know what to look for.
The submission exists before anyone reads the emailNo attachment to save by hand. The pack lands in one case, typed, in the property queue for the Great Lakes region.
Step 2AI check

The values and the construction come out of the pack

The AI reads the schedule of values and the application, and writes the figures into the submission: two sites, the foundry at Parma for $24.6M and the machining plant at Solon for $13.8M, plus the construction, the sprinklers and the business interruption sum.

One value is missing. The application says insulated metal panels on site 1 and never states their core, which matters on a site with hot work.

Risk detailsCase management
Each figure names the document it was read from.
Nobody retypes a schedule of valuesThe figures are in the submission fields within minutes, each one traceable to its file.
Step 3AI check

The pack is checked against the guidelines

Your guidelines for metalworking risks ask for five years of loss runs, a survey less than three years old and the panel core on hot work sites. The underwriting manager wrote them in plain language. An AI agent runs the pack against them and writes a status and its reasoning onto the submission.

Five years of loss runs?Survey under three years old?Panel core stated?

Here the status is Attention required: the loss runs start in 2023, the survey dates from March 2023, and the panel core is not stated.

AI analysis resultsAI console
Every warning names the guideline and the document it was checked on.
The gaps are known on day oneMegan reads three reasons before she opens a single attachment. The broker can be asked while ten days remain.
03

The missing pieces

Twelve submissions on Megan's list this week. Which ones first, and what does each broker still owe her?

Step 4Assistant

The week's submissions are sorted, and the authority question answered

Megan's queue shows each submission with its verdict, its values and the date the broker needs an answer. Ridgeline Castings is the largest. She asks the assistant whether she can quote it within her authority.

Ask about this submissionAI console
The answer cites the authority grid and the guidelines, both filed in the repository.
The referral is known before the pricing startsMegan plans the manager's approval into her week, rather than learning about it the day the quote is due.
Step 5Assistant

The request to the broker is drafted from the pack

Megan asks the assistant for a request to Mark. The draft lists the 2021 and 2022 loss runs and the panel core material, with the client name and the quote date already in it. She adds an offer to visit the site with the risk engineer, and sends it. The step carries a three-day timer.

Submission SUB-26-1187Case management
The request and the broker's reply stay in the submission, with their dates.
The broker hears about it the same morningA precise request the day after the pack arrives, with the reason. The broker has time to ask the client.
Step 6Viewer

The survey is read for its open recommendations

While the broker gathers the documents, Megan opens the 2023 survey in the viewer, next to the submission. She asks which recommendations are still open, and the answer points to the pages. Julie Bennett, the risk engineer, opens the same page and leaves a note on R2.

Survey March 2023.pdfViewer
Julie's note sits on the page, visible to whoever opens the survey next.
Look at your last renewals: how often did a recommendation from the survey reach the price, and how often did it stay on page 14?
The survey's open points reach the priceThe underwriter and the risk engineer work on the same page, and the note stays there for the next renewal.
04

The risk and the quote

The broker's reply arrives on Thursday. The price is Megan's to set, inside the limits her manager wrote. Three steps take it to the broker.

Step 7AI check

Five years of losses, and the check runs again

The 2021 and 2022 loss runs and the broker's answer on the panels come in by email and are filed on the same submission. The AI extracts each loss with its year, cause and amount, the check runs again, and the status moves to Complete.

Loss historyCase management
The earlier warning stays in the history, so the trail shows how it was cleared.
The large loss and the open point line upThe 2022 fire and recommendation R2 concern the same ducts. Megan sees it on one screen before she prices.
Step 8Workflow

The terms are set and the quote goes for approval

Megan enters a rate of $0.118 per $100 of insured values, a $25,000 deductible, and a $100,000 deductible for fire in the extraction system until R2 is closed. Three conditions come with the quote, two of them before inception, and each becomes a task on the submission. Because the values exceed $30M, the approval path sends the quote to David Mercer, who approves the same afternoon.

Authority: $25M per siteTotal: $38.4M
Quote QTE-26-0942Case management
Each condition is a task with its owner and its due date.
The manager approves on the full fileNo email with nine attachments to forward. The approval travels with the submission and everything in it.
Step 9Assistant

The quote goes out with a letter that explains it

The quote document comes from your rating tool through the API and is filed on the submission. Megan asks the assistant for the cover letter. The draft explains the premium, the two deductibles and the conditions, with their reason in the survey. She softens one sentence and sends it on Nov 18, two days early.

Cover letterAI console
The values in the draft come from the submission, and each one can be checked.
The broker can place a quote that explains itselfEvery term carries its reason and its source, which helps the broker now and the next underwriter at renewal.
05

The verdicts

The AI never sets a premium. It reads the pack against your guidelines and names its state, and that state decides whether Megan prices, asks or declines.

Complete

Every guideline passes

Activity, values, loss history, survey, protection. One line on the submission records that each guideline held.

Next: the underwriter reads the risk and prices it.
Attention required

A piece is missing

Loss runs too short, a survey too old, a value not stated. The analysis says which, with its source.

Next: the broker gets a precise request; the check runs again when the reply arrives.
Not eligible

Outside appetite, and you know why

The AI lists what fits first, then names the guideline the risk falls outside.

Next: a reasoned decline, written and signed by the underwriter.
The limits we set

It sets no rate, accepts no risk and declines no broker. Across the nine steps on this page, the price and the terms are the underwriter's, and the approval above her authority is her manager's.

The AI reads, extracts, checks and drafts. The control is in how the guidelines and the approval path are written, so it holds in the busiest renewal season.

06

In numbers

Counted on Ridgeline Castings, one renewal over ten days. A pack with fifteen attachments or a stricter guideline will count differently.

0 valuesretyped from the schedule of values. The sites, sums and construction come out of the pack.
Day 2when the broker is asked for the missing loss runs, with ten days left before the quote.
1 approvalabove authority, routed by the approval path and taken on the full file.

Where these numbers come from. They are counted on the scenario described on this page, with example data. They are not a customer result.

Why these numbers hold

Zero wasted readingProductivityThe pack is typed, extracted and checked before the underwriter opens it. She starts from the gaps.
Zero lost attachmentsProcess reliabilityRequests, replies, notes and approvals live on the submission. Nothing sits in a mailbox.
Decision supportA person decidesAnswers and drafts cite their sources. The rate, the terms and the approval stay with people.
07

Built for

The underwriter, her manager and the CIO each need a reason before a property team works this way.

Business operations

Commercial lines underwriters

Most of the week goes into opening attachments, and the risk gets what time is left.

  • The pack arrives typed, with the values in the submission.
  • The gaps are named against your own guidelines.
  • The request to the broker is drafted with the right references.
  • The survey opens beside the submission, and answers cite their page.
Risk and compliance

Underwriting managers

Quotes reach you for approval with no sign of how the guidelines were applied.

  • The AI prices nothing. The terms are the underwriter's.
  • Each verdict names the guideline it checked.
  • Approvals above authority follow the path you wrote.
  • Every request, reply and approval is dated on the submission.
IT and architecture

CIOs and architects

The brokers' mailbox, the rating tool and the policy system have never been introduced.

  • Broker emails and attachments land in the submission case.
  • The rating tool files its quote document through the API.
  • The guideline check runs as a step of the submission process.
  • Hosted on your servers or in your cloud, as your policy requires.

How to begin

With one guideline set. Take the checklist your underwriters run by hand on every property pack, and move it to the door. A season of renewals is enough to judge it.

Your brokers' mailbox, your rating tool and your policy system stay where they are. The check sits between them.

Next step

Try it on one of your submissions

Bring an anonymized submission pack and your underwriting guidelines. Watch the check run on your own documents, then decide on the price yourself.

The screens on this page are drawn to show one step at a time, with example data.

Uxopian Software helps organizations put AI to work across the systems they already run, without giving up control of their data, their infrastructure or their critical processes.