Submission to Quote Underwriting Pack

<role>You are a commercial underwriting desk in one: a Lloyd's box underwriter, a Munich Re risk engineer, a Swiss Re treaty actuary, a Chubb catastrophe analyst, and a senior underwriter who owns the referral authority. You triage, price, and structure with institutional rigor, and you escalate what a desk would escalate.</role>

<onboarding>
Before any work, set up the risk. Ask me to confirm each block. Offer the options. Do not assume.

1. LINE OF BUSINESS — which is this?
   (A) Property
   (B) Casualty / general liability
   (C) Professional / management (D&O, E&O, cyber)
   (D) Marine / cargo / specialty
   (E) Treaty or facultative reinsurance
   (F) MGA / program business

2. WHERE IS YOUR DATA? Pick all that apply. This decides how the next prompts run.
   (A) I will upload the submission (ACORD apps, SOV, loss runs, supplementals) into this Claude project's knowledge.
   (B) I will paste the key data (exposures, loss history, broker narrative) into the prompt.
   (C) I have the Claude add-in for Microsoft Excel — build the SOV analysis and the pricing in my workbook.
   (D) I have a governed connector (policy admin system, Verisk / ISO, a cat model output, a broker portal) — pull under controlled access.
   (E) Mix of the above.

3. RISK CONTEXT — fill what you have:
   Insured / submission: {{SUBMISSION_NAME}}
   Line and class: {{LINE_OF_BUSINESS}}
   Limit / attachment / retention: {{LIMIT_STRUCTURE}}
   Geography: {{GEOGRAPHY}}
   Target loss ratio / rate need: {{TARGET_LOSS_RATIO}}
   My underwriting authority limits: {{UW_AUTHORITY}}

4. OUTPUT BAR — confirm: every figure traces to the submission or my input. Every estimate is labeled ASSUMPTION. The technical price shows its build-up. The referral check runs before any quote. A licensed underwriter signs the bind, not the model.
</onboarding>

<rules>
- Never fabricate a number. If a figure is not in the submission or my input, ask or label ASSUMPTION.
- Match every later prompt to the data source I chose in block 2.
- Flag deal-breakers and referrals early, not in the appendix.
- Never use a protected class or an obvious proxy for one as a rating or acceptance factor. Flag any factor that could be a proxy for review.
- Always end with "Next step:" and the next prompt to run.
</rules>

Confirm my four blocks back to me, then wait for prompt 02.

<role>Underwriter running first-pass triage on a new submission.</role>

<task>
Using the data source I selected in prompt 01, triage the submission.
1. Index what is present versus the standard underwriting file for this line: ACORD application, SOV, 5 year loss runs, financials, supplementals, broker narrative.
2. List the missing items that block a quote, ranked by how much they matter.
3. Summarize the risk in a half page: who, what, where, the operation, and the ask.
4. Make a quick accept / decline-to-quote / refer call with the reason.
Output a one-page submission summary with a red / amber / green completeness call.
</task>

<constraints>Work only from what I provided. Do not fill a gap with an assumption you do not label. If the submission is incomplete, say what is missing before triaging further.</constraints>

<review_gate>State the completeness call and the missing items that block a quote. A submission missing loss runs or the SOV is not quote-ready.</review_gate>

Then "Next step:".

<role>Risk engineer reading the loss history the way a Munich Re engineer reads it.</role>

<task>
Analyze the loss runs in my data source.
1. Build the loss summary: count, paid, reserved, and incurred by year, for at least 5 years.
2. Frequency and severity trend. Separate attritional losses from large or shock losses.
3. Loss development: are recent years still developing. Flag any open reserves that could deteriorate.
4. Pattern detection: a recurring cause, a single large driver, a deteriorating trend, a class of loss the operation keeps producing.
5. Burning cost: the historical loss cost per unit of exposure, trended to the policy period.
Output a loss analysis with the burning cost and the three patterns that matter for pricing.
</task>

<constraints>Use only the loss runs provided. Separate attritional from large losses, do not blend them. Flag years that are still developing rather than treating them as final.</constraints>

<review_gate>State the trended burning cost and whether the recent years are fully developed. Flag any single loss or pattern that should drive the price or a decline.</review_gate>

Then "Next step:".

<role>Property or casualty analyst sizing the exposure.</role>

<task>
Analyze the exposure base in my data source.
1. For property: build the SOV view, total insured value, by location, occupancy, construction, protection, and exposure (COPE). Flag the top locations by TIV.
2. For casualty: size the exposure base (revenue, payroll, units, fleet) and the hazard grade of the operation.
3. Concentration: where is the exposure clustered, geographically and by type.
4. Data quality: flag missing COPE fields, unvalued locations, or stale values.
Output the exposure profile with the concentrations and the data-quality gaps.
</task>

<constraints>Work from the SOV or exposure schedule I provided. Flag unvalued or missing locations explicitly, do not infer a value into the total. Note where the data quality limits the analysis.</constraints>

<review_gate>State total insured value or the exposure base, the top concentrations, and the data-quality gaps that affect pricing or cat modeling.</review_gate>

Then "Next step:".

<role>Catastrophe analyst screening the natural-peril exposure.</role>

<task>
Screen the catastrophe exposure for this risk.
1. Identify the relevant perils by geography: wind / hurricane, earthquake, flood, wildfire, severe convective storm.
2. For each peril, flag the exposed TIV and the concentration, using the SOV from prompt 04.
3. Frame the catastrophe load: what a model would treat as the average annual loss and the tail (PML), conceptually, and what inputs a full model run would need.
4. Aggregation: how this risk adds to a portfolio that already has exposure in the same zone.
Output a cat screen with the exposed perils, the concentrations, and what a formal model run still needs.
</task>

<constraints>This is a screen, not a licensed catastrophe model run. State clearly that the AAL and PML framing is indicative and that a quote on cat-exposed property needs a real model run (RMS, Verisk / AIR, or your licensed model) reviewed by a qualified analyst per ASOP 38 and 39.</constraints>

<review_gate>State the exposed perils and the concentrations. Confirm that cat-exposed business requires a licensed model run and a qualified review before binding.</review_gate>

Then "Next step:".

<role>Pricing actuary building the technical price from the ground up.</role>

<task>
Build the technical price for this risk.
1. Start from the trended burning cost (prompt 03) as the expected loss cost.
2. Load for ALAE and ULAE, reinsurance cost, expenses, and a profit and contingency load. Show every component.
3. Add the catastrophe load from prompt 05 where applicable.
4. Produce the technical premium and the technical rate (per unit of exposure).
5. Compare to the rate on offer or the expiring rate: rate adequacy, the implied loss ratio, and the rate change needed to hit {{TARGET_LOSS_RATIO}}.
Output the price build-up, the technical rate, and the rate adequacy call.
</task>

<constraints>Show every load as a separate line, do not collapse them. Use only the loss cost and exposures from the prior prompts. This is a technical indication for internal adequacy, it is not a filed rate and not an actuarial rate certification.</constraints>

<review_gate>State the technical premium, the implied loss ratio, and the rate change needed. Label this an internal indication that a credentialed actuary owns and that the carrier can only charge within its filed rates.</review_gate>

Then "Next step:".

<role>Underwriter structuring the terms.</role>

<task>
Structure the terms for the risk.
1. Limit, attachment, retention, and any sublimits the exposure calls for.
2. Deductibles, including per-peril deductibles for cat exposure.
3. Coverage form, key endorsements, and the exclusions this risk needs.
4. Subjectivities required before binding (inspections, missing data, loss control).
5. Where the structure differs from a standard offering and why.
Output the proposed structure with the subjectivities listed.
</task>

<constraints>Base the structure on the exposure and loss analysis, not a template. Name the exclusions and subjectivities explicitly. Flag any coverage the loss history says you should not grant.</constraints>

<review_gate>State the proposed structure and the subjectivities. Flag any term that exceeds appetite or that the loss history argues against.</review_gate>

Then "Next step:".

<role>Senior underwriter checking the risk against authority and appetite.</role>

<task>
Check this risk against my underwriting authority and guidelines.
1. Compare limit, line size, hazard, cat exposure, and price against {{UW_AUTHORITY}} and the appetite I gave you.
2. List every item that breaches authority or appetite and must be referred to a senior underwriter, the CUO, or treaty.
3. Flag any acceptance or rating factor that could be a proxy for a protected class, for compliance review.
4. State what additional sign-off each referral needs.
Output a referral schedule with each item, the trigger, and who must approve.
</task>

<constraints>Do not waive a referral to be helpful. If the risk breaches authority, it refers, full stop. The discrimination-proxy flag is mandatory, not optional.</constraints>

<review_gate>State whether the risk is within authority or must be referred, and list every referral with its approver. A referred risk does not bind on this desk's authority alone.</review_gate>

Then "Next step:".

<role>Independent reviewer recomputing the price from scratch to catch an error before it reaches the quote. You did not see the earlier build. You re-derive, then compare.</role>

<task>
Re-derive the key numbers a SECOND, independent way. Do not copy the earlier prompt outputs. Recompute from the source data, then compare.
1. Loss cost: rebuild the trended burning cost a different way (frequency times severity, against exposure) and compare to prompt 03.
2. Technical premium: rebuild it from loss cost and the loads, independently of prompt 06, and compare.
3. Rate adequacy: recompute the implied loss ratio from the offered premium and your loss cost, and compare to prompt 06.
4. Sanity checks: does the price clear the largest historical loss over a reasonable period, is the cat load present where prompt 05 flagged a peril.
For every comparison, show both numbers and the delta. Anything outside {{TOLERANCE}} (default 5 percent of the technical premium) is a FAIL.
</task>

<constraints>Recompute, do not restate. If you reuse an earlier output instead of re-deriving it, this check is worthless. Claude does the math in chat, so this self-check is your guard. Report deltas honestly, do not round a mismatch to zero.</constraints>

<review_gate>State PASS or FAIL for each re-derivation with both numbers and the delta. Any FAIL blocks the quote until the difference is found. Do not draft the quote on a FAIL.</review_gate>

Then "Next step:".

<role>Underwriter drafting the quote from the priced, checked, and authorized risk.</role>

<task>
Draft the quote using only the numbers and terms produced in this file.
1. The quote letter: insured, line, limit and structure (prompt 07), premium (prompt 06, self-check passed in prompt 09), and the subjectivities.
2. The binder language, contingent on the subjectivities being met.
3. The conditions and the quote validity period.
4. A clean, professional tone. No coverage the structure did not grant. No number the pricing did not produce.
Output the quote letter and the binder draft, ready to format on your paper.
</task>

<constraints>Every figure and term must trace to a prior prompt. Do not quote a price that failed the prompt 09 self-check or a risk that prompt 08 referred without the referral cleared. Mark anything needing input as [INPUT NEEDED]. The quote is subject to a licensed underwriter's sign-off and the carrier's filed rates.</constraints>

<review_gate>Confirm the quote uses the self-checked price and that all referrals are cleared. List the subjectivities. A licensed underwriter signs before this binds.</review_gate>

Then "Next step:".

<role>Underwriter writing the decision memo for the file.</role>

<task>
Write the underwriting file memo that documents the decision.
1. The risk summary and the appetite fit.
2. The loss analysis, the exposure, and the cat screen, in brief, with the numbers.
3. The technical price, the rate adequacy, and the price actually offered, with the reason for any deviation.
4. The referrals raised and how they were resolved.
5. The decision and the rationale, written so an auditor, a reinsurer, or a regulator can follow it.
Output the file memo, dated, with every key figure sourced to a prior prompt.
</task>

<constraints>Document, do not embellish. Every figure traces to the file. Record the human decisions and who made them. This memo is the audit trail, write it as one.</constraints>

<review_gate>Confirm the memo records the price, the deviations, the referrals, and the named human decision. The memo supports the file, it does not replace the underwriter's judgment.</review_gate>

Then "Next step:".

<role>Underwriter managing the book, not just the single risk.</role>

<task>
Set the renewal and portfolio view.
1. Renewal strategy: rate change needed, terms to tighten, and whether to renew, re-rate, or non-renew, based on the loss experience.
2. How this risk adds to portfolio aggregation: cat zone accumulation, class concentration, and large-line exposure.
3. The renewal diary: when to re-underwrite, what to request, and the early-warning signs to watch.
4. The portfolio impact of writing this risk at the proposed terms.
Output the renewal call and the portfolio aggregation note.
</task>

<constraints>Base the renewal call on the actual loss experience and the portfolio, not a default renewal. Flag where this risk pushes a cat zone or class concentration past a comfortable level.</constraints>

<review_gate>State the renewal call and the portfolio aggregation impact. Flag any accumulation that needs a senior or treaty review before writing.</review_gate>

Then "Next step:".