Prompt Chain: Submission-to-Renewal Decision in Under 30 Minutes

What This Builds

You'll build a 4-step prompt chain — a saved sequence of AI prompts that takes a new submission or renewal from initial review through underwriting analysis, coverage decision, and agent communication in under 30 minutes. Each step is saved in a reference document so you can run the same process consistently on any account without reinventing the approach.

Prerequisites

• Comfortable using Claude Pro or ChatGPT Plus for submission analysis (Level 3)
• A Word or Google Docs document to save your prompt chain (reusable reference)
• An active submission or past account to test on
• Cost: Free (Claude Pro or ChatGPT Plus recommended)

The Concept

A prompt chain is a pre-defined sequence of AI interactions where each step builds on the previous one. Instead of starting fresh for each account and reinventing your analysis approach, you follow the same proven sequence: analyze → decide → document → communicate. The chain gets faster with each use as you refine the prompts for your typical account types.

Think of it as a systematic underwriting workflow — but AI-assisted:

• Step 1 = Submission Analysis (what is this risk?)
• Step 2 = Coverage Decision (quote / decline / refer — and on what terms?)
• Step 3 = File Note (document the decision)
• Step 4 = Agent Communication (deliver the decision)

Build It Step by Step

Part 1: Create Your Prompt Chain Document

Open a new Google Doc or Word document. Name it "Underwriting Prompt Chain — [Your Name/Line]."

Create 4 sections. Copy the prompts below into each section, then customize them for your book.

Part 2: Write and Save Each Prompt

Step 1: Submission Analysis

Copy into your document under "Step 1":

I'm a commercial insurance underwriter reviewing a submission. Analyze the following and give me:

1. Business description: what they do, operations, years in business, size (revenue/employees)
2. Coverage requested: lines, limits, deductibles
3. Key risk factors: top 3–5 factors that will drive my underwriting decision (positive and negative)
4. Loss history assessment: frequency/severity trend, any patterns or large outlier losses
5. Missing information: what is absent that I need before making a decision?
6. Preliminary risk quality: above-average / average / below-average — and brief reasoning

Submission details: [paste submission summary or text]

Step 2: Coverage Decision Framework

Copy into your document under "Step 2":

Based on the submission analysis above, help me think through my coverage decision.

My parameters for this account type:
- Quote authority: [e.g., premium up to $75K / TIV up to $15M]
- Appetite criteria: [e.g., preferred if 5-year loss ratio under 65% / avoid habitational over 4 stories]
- Standard conditions for this class: [e.g., assault and battery exclusion for hospitality / mold exclusion for older buildings]
- Referral triggers: [e.g., prior losses over $50K / unusual operations / premium over my authority]

For this account, recommend:
1. Quote, decline, or refer — and the primary reason
2. If quoting: what conditions or endorsements should apply?
3. If declining: what is the primary decline reason to cite to the agent?
4. If referring: what additional information should I provide in the referral?
5. What is the one thing I most need to verify before finalizing my decision?

Step 3: Underwriting File Note

Copy into your document under "Step 3":

Draft a professional underwriting file note documenting my review and decision on this account. The note will be saved in the policy system.

Decision: [quote / decline / refer / renew with changes]
Key underwriting factors that drove the decision: [the 2–3 most important positive or negative factors]
Coverage terms: [any endorsements, conditions, or exclusions applied]
Loss history summary: [brief summary — number of claims, total losses, trend]
Any follow-up required: [information still needed, items to verify, conditions precedent to binding]

Write in past tense, factual tone, 150–250 words. No subjective language about the agent or insured.

Step 4: Agent Communication

Copy into your document under "Step 4":

Draft a professional email to the agent communicating my underwriting decision.

Decision: [quote / decline / conditional acceptance / renewal terms]
Agent name: [first name]
Key information to communicate: [premium / conditions applied / reasons for decline / information still needed]
Tone: [professional and relationship-preserving / direct and factual]
Any deadline: [respond by date if relevant]

For a quote: include what I need from them to bind (signed application, payment, etc.)
For a decline: be factual about the reason; close with an invitation to bring other business.
For a renewal with conditions: explain what is changing and why, in plain language.
Include a subject line.

Part 3: Run the Chain on a Test Account

Pick a recent account (won, lost, or current) and run through all 4 steps in a single Claude or ChatGPT conversation. Keep all steps in the same conversation — the chain works because context builds.

For each step:

1. Open your prompt chain document
2. Copy the Step [X] prompt
3. Fill in the bracketed fields with the account information
4. Paste into Claude/ChatGPT
5. Read the output — use it as input for the next step

Total runtime for a typical account: 20–30 minutes.

Real Example: Full Chain in Action

Starting situation: A commercial general liability renewal arrives for a regional plumbing contractor, 12 employees, $1.8M revenue. Their current policy expires in 45 days. Loss history shows 2 GL claims in 5 years totaling $22,000.

Step 1 (Day 1, 9am) — Submission Analysis: Input: Account description, current policy terms, 5-year loss run Output: Risk profile summary — medium-sized contractor, clean loss history for the class, frequency acceptable, 2 claims are routine for a plumbing contractor with this revenue, no red flags. Missing: updated certificate of insurance for subcontractors used.

Step 2 (9:10am) — Coverage Decision: Input: Analysis from Step 1, my appetite parameters (preferred contractor risk, within my authority, standard GL conditions apply) Output: Recommend quoting renewal at 5% increase (loss history justifies modest increase; market conditions flat). Conditions: subcontractor certificate requirement remains; add completed operations sublimit per prior. Request subcontractor list before binding.

Step 3 (9:20am) — File Note: Input: Decision and key factors from Step 2 Output: Professional file note documenting renewal decision, 200 words, ready to paste into policy system.

Step 4 (9:25am) — Agent Email: Input: Decision — renewing with 5% increase and existing conditions; need subcontractor certs before binding Output: Professional email to agent delivering renewal terms, explaining the modest increase, requesting the cert before binding, and noting the 45-day timeline.

9:30am: Account fully analyzed, documented, and communicated in 30 minutes.

What to Do When It Breaks

• Step 2 decision framework doesn't match my company's actual appetite → Your Step 2 parameters are too generic. Add 3–5 specific examples from your book: "For GL contractors, we require completed operations to equal GL limits; we decline accounts with any prior products liability claims over $25K."
• File note is too long or too short → Adjust the word count in Step 3: "150 words maximum — brevity is valued in file notes" or "250 words minimum — include more detail on the loss history."
• Step 4 email doesn't match your firm's tone → Add your firm's typical email style: "Our agents are regional, relationship-focused. Emails should be warm but professional — first-name basis with established agents, formal with new agents."
• Chain loses context between steps → Start Steps 3 and 4 with: "We're continuing the underwriting analysis from this session. The account is [brief description]. The decision is [decision]. Now..." to refresh context before the step.

Variations

• Simpler version: Run just Steps 1–2 (analysis + decision framework) for every account; write Steps 3–4 manually. The chain's biggest time savings are in the analysis and decision steps.
• Renewal surge version: During January 1 renewal season, run Step 1 on 10 accounts in parallel (one account per conversation) across multiple browser tabs, then process through Steps 2–4 for each. Dramatically accelerates the renewal sprint.
• Referral version: Add a Step 2b: "Draft the referral submission to my regional manager" — the AI writes the referral memo based on the Step 2 decision framework output.

What to Do Next

• This week: Test the full chain on your most recent lost renewal — see what the AI would have produced and whether the decision and communication would have been better
• This month: Refine the prompts after running the chain on 10 real accounts — the bracketed fill-in areas get tighter and more effective with each use
• Advanced: Combine this chain with the Claude Project content guide — Step 1 pulls from your appetite guidelines document; Step 4 drafts use your firm's standard letter templates

Advanced guide for Insurance Underwriter professionals. All underwriting decisions require professional judgment and appropriate authority approval.