Your First Prompts for Insurance Work

Prompts are the steering wheel of AI. The same model can produce a half-baked email or a polished, claims-grade communication depending on how you ask. This lesson teaches you a simple framework you can use the rest of your career.

What You'll Learn

• A reliable prompt structure tailored to insurance tasks
• How to use roles, context, and constraints to improve output
• Real prompts you can copy for daily insurance work
• How to iterate when the first answer is not good enough

The C.L.A.I.M. Framework

Most beginner prompts fail because they give the model nothing to anchor on. To fix that, use C.L.A.I.M. — five elements every good insurance prompt includes:

• Context: the line of business, jurisdiction, parties involved, and any background facts
• Line of work: what role is the AI playing (claims adjuster, broker, underwriter)
• Action: the specific task you want done (summarize, draft, analyze, compare)
• Inputs: the documents, facts, or numbers the AI needs
• Must-haves: format, length, tone, exclusions, and required disclaimers

A weak prompt: "Write a claim denial letter."

A C.L.A.I.M. prompt:

Context: I am a senior claims adjuster handling a homeowners
property claim in Texas. The insured filed for water damage.
Investigation showed the loss was caused by a long-term
plumbing leak that pre-existed the policy.

Line of work: Act as an experienced HO-3 claims adjuster.

Action: Draft a denial letter to the insured.

Inputs:
- Policy: HO-3
- Claim number: HO-2026-44812
- Date of loss: 2026-03-15
- Cause: long-term seepage from supply line under the kitchen sink
- Exclusion: ISO HO 00 03 05 11, Section I — Exclusions, A.10
  (constant or repeated seepage of water over a period of weeks)

Must-haves:
- Empathetic but clear tone
- Quote the relevant exclusion language but do not invent
  any policy text I have not provided
- Include standard appeal rights language placeholder
- Under 350 words
- US English
- Add a placeholder for state-required notices

That second prompt produces a usable first draft. The first prompt produces a generic letter you would still need to rewrite.

The Three Most Useful Prompt Patterns in Insurance

1. The Summarization Prompt

You are a commercial lines underwriter. I am pasting a 12-page
property submission packet. Summarize it in three parts:

1. The five most important facts about this risk
2. Three underwriting concerns I should investigate
3. Three questions I should ask the broker before quoting

Limit each item to one sentence. Do not invent any details
not present in the document.

Document:
[paste the submission text here]

2. The Drafting Prompt

You are a friendly, professional insurance broker writing to a
small business client. Draft a renewal email for ABC Bakery
covering BOP renewal. Premium increased 11 percent due to
hardening market conditions in casualty. Tone: warm,
transparent, no jargon. Length: 180-220 words.
End with a CTA to schedule a 15-minute review call.

3. The Analysis Prompt

You are an experienced claims supervisor. I am pasting two
versions of an adjuster narrative on the same auto bodily
injury claim. Compare them. List:

- Three factual discrepancies between the two narratives
- Two coverage issues that may be missing from both
- One recommendation for next steps

Narrative A: [paste]
Narrative B: [paste]

Tips for Better Insurance Prompts

Always specify the line of business and jurisdiction. "Auto liability in California" gets you very different output than "homeowners in Florida." The legal and regulatory environment matters.

Tell the AI what NOT to do. Phrases like "do not invent policy form numbers," "do not give legal advice," or "do not name a specific carrier" sharpen the output.

Give a target reading level. "Plain English at a 7th grade reading level" is a useful instruction for any consumer-facing communication. Many state DOIs require plain language disclosures.

Constrain length and format. "Three bullet points," "under 250 words," "in a table with these columns" — the model will respect these constraints.

Iterate, do not start over. If the first draft is 80 percent there, follow up with "tighten the second paragraph and remove the legal jargon" rather than rewriting your prompt from scratch.

Prompts to Try Today

Copy and adapt these:

Act as a personal lines insurance broker. A new client just
emailed asking the difference between actual cash value and
replacement cost on their HO-3 dwelling coverage. Write a 3-paragraph
reply in plain English, no acronyms, friendly tone.

Act as a workers' compensation claims examiner. Summarize this
medical report in 5 bullet points: diagnosis, treatment plan,
work restrictions, expected MMI date, and any red flags.
Do not invent any clinical details.

Report: [paste the medical narrative]

Act as a commercial lines underwriter. List 8 questions I should
ask before quoting a manufacturing risk with $4M in revenue,
flammable solvents on premises, and a 5-year prior carrier
history I have not yet seen.

Key Takeaways

• The C.L.A.I.M. framework (Context, Line of work, Action, Inputs, Must-haves) consistently produces stronger insurance prompts.
• The three most useful patterns are summarization, drafting, and analysis. Almost every insurance task fits one of them.
• Always specify line of business and jurisdiction. Always tell the AI what NOT to do.
• Iterate on the draft. Do not abandon a prompt at the first attempt.