Prompt Library for Adjuster Reports and Estimates

Writing adjuster reports is a significant part of the job, and it’s some of the least satisfying work: long, repetitive, and requiring consistent language to hold up under audit. Every field inspection produces a stack of notes — measurements, photos, observed conditions — that needs to be turned into a structured report that coverage supervisors, attorneys, and re-insurers can actually use.

AI is surprisingly good at this. Not because it understands insurance, but because adjuster reports follow predictable patterns. Damage description, causation analysis, scope of loss, coverage analysis, reserve recommendation. The structure barely changes claim to claim. What changes is the specific facts, and you’re supplying those.

Where AI fails is the same place it fails in most professional contexts: when it doesn’t know something, it fills in the blank with plausible-sounding fabrications. It will state building codes it made up, reference carrier policy language it’s never seen, or describe a damage mechanism that doesn’t match what you observed. Every output needs a real review.

This prompt library is what I actually use day-to-day. It covers the most common report types, with exact prompts, example outputs, and notes on what to watch for before you submit.

Prompt 1: Initial Field Inspection Summary

When to use: Right after a site visit, before writing the full report. You’ve got handwritten notes and a photo set. This gets you a clean summary that becomes the narrative section of your report.

The prompt:

You are an experienced property insurance adjuster writing an inspection summary.

Claim type: [Water damage / Fire / Hail / Wind / etc.]
Loss date: [Date]
Inspection date: [Date]
Property type: [Single-family residential / Commercial / etc.]

Observations from field notes:
[Paste your raw notes here — room by room, measurements, materials, conditions observed]

Write a field inspection summary in professional adjuster language. Format: overview paragraph, then room-by-room findings using bullet points. Note: do not assume causation or coverage. Only describe what was observed. Avoid passive voice where possible.

Example output (paraphrased): The model produced a clean two-paragraph overview followed by room-by-room bullets. Tone was professional without being verbose. It correctly avoided speculating on causation when the prompt told it not to.

Watch for: The model sometimes upgrades observed conditions to seem worse. “Staining on drywall” became “significant water infiltration damage” in one test. Always match the adjective intensity against your actual notes.

Prompt 2: Causation Analysis Paragraph

When to use: When your report needs an explanation of how the damage occurred — the critical section that affects coverage determination.

The prompt:

Write a causation analysis paragraph for an insurance claim report.

Claim type: [Water / Fire / Wind / Hail]
Mechanism of loss: [e.g., "supply line failure at kitchen sink shutoff valve, slow leak estimated 3-6 weeks based on staining pattern and dried efflorescence"]
Policy type: [HO-3 / DP-1 / Commercial property]

Write a 2-3 paragraph causation analysis in standard adjuster report language. State the probable cause of loss, supporting evidence from inspection, and any contributing conditions. Do not make coverage determinations. Note any areas requiring further investigation (e.g., "recommend plumber evaluation of supply line origin to confirm single event vs. slow leak").

Example output (paraphrased): Solid structure. The model correctly separated observed evidence from probable cause and flagged the single-event vs. slow-leak distinction as a coverage-relevant question requiring documentation.

Watch for: If the prompt includes a word like “gradual,” the model will often write language that sounds like it’s heading toward a coverage exclusion even when that’s not your job. Keep causation analysis factual and neutral.

Prompt 3: Scope of Loss Narrative

When to use: When you need to explain the scope of repairs in plain language before the Xactimate estimate. This goes in the body of the report, not the estimate itself.

The prompt:

Write a scope of loss narrative for an insurance claim.

Damaged area: [e.g., "kitchen, adjacent dining room, and hallway — approximately 280 sq ft affected"]
Materials affected: [e.g., "LVP flooring, drywall to 4 feet, base cabinets on one wall, insulation in crawlspace below"]
Required work: [e.g., "Demolition of affected flooring and drywall, antimicrobial treatment, structural drying, replacement of LVP and drywall, reinstall base cabinets, repaint affected walls"]

Write a 2-paragraph scope of loss narrative in adjuster report language. Include demolition, mitigation, and repair phases in logical order. Do not include line item pricing. Keep the language direct and specific.

Example output (paraphrased): The output organized demolition, mitigation, and repair into a logical sequence with appropriate professional language. No puffery. The paragraphs were about 80 words each — appropriate length.

Watch for: The model sometimes adds “industry-standard drying protocols” without specificity. If IICRC S500 standards are relevant to your claim, name them explicitly in your prompt, otherwise the model will use vague language that won’t satisfy a carrier’s technical reviewer.

Prompt 4: Coverage Analysis Summary

When to use: When you need to document your coverage analysis — what applies, what doesn’t, and why. This is the section with the most liability if you get it wrong, so treat AI output as a first draft only.

The prompt:

Write a coverage analysis summary for an insurance claim.

Policy type: [HO-3 / commercial / etc.]
Applicable coverage: [e.g., "Coverage A — Dwelling: sudden and accidental water damage from supply line failure"]
Applicable exclusions to note: [e.g., "Long-term seepage exclusion may apply if slow leak is confirmed — see causation analysis"]
Additional coverages triggered: [e.g., "Coverage D — Loss of Use: insured displaced for duration of repairs"]

Write a coverage analysis summary paragraph (150-200 words) in adjuster report language. State what coverage applies and reference the policy sections. Note any coverage questions requiring supervisor review. Do not make final coverage determinations — frame as preliminary findings pending review.

Example output (paraphrased): The model correctly framed the analysis as preliminary and flagged the slow-leak question for escalation. Language was appropriately hedged without being unclear.

Watch for: Do not paste actual policy language into this prompt. AI will try to interpret it and may misstate what it says. Write your own coverage analysis based on your policy reading; use AI only to format and structure what you already know.

Prompt 5: Reserve Recommendation Memo

When to use: When you need to document your recommended reserve for supervisor review.

The prompt:

Write a reserve recommendation memo for an insurance claim.

Claim number: [Number]
Date of loss: [Date]
Recommended reserve breakdown:
  - Dwelling: $[amount]
  - Contents: $[amount]
  - ALE/Loss of Use: $[amount]
  - Subrogation potential: [Yes/No — brief note]
Basis for recommendation: [e.g., "Xactimate estimate attached; ALE based on 6-week repair timeline at comparable rental rate of $1,800/month"]

Write a 150-word reserve recommendation memo in professional claims language. Note the basis for each reserve component and flag any items requiring further development.

Example output (paraphrased): Clean, well-structured. Each reserve component had its own sentence with the stated basis. The flagged items section was useful and appropriate.

Watch for: Reserve amounts in any AI output should never be treated as recommendations — AI has no idea what your market rates are. Only use this prompt to format reserve amounts you have already calculated.

Prompt 6: Subrogation Potential Evaluation Note

When to use: When you need to document whether the claim has subrogation potential and why.

The prompt:

Write a subrogation evaluation note for a claim file.

Cause of loss: [e.g., "Supply line failure — copper pipe with visible manufacturing defect at compression fitting"]
Potentially responsible party: [e.g., "Pipe manufacturer / original plumber who installed fitting"]
Evidence collected: [e.g., "Retained failed pipe section, photos of fitting, original installation date from permit records"]
Recommended action: [e.g., "Preserve evidence, refer to SIU for subrogation evaluation before paying claim"]

Write a 150-word subrogation potential note in claims file language. State the basis for subrogation potential, evidence status, and recommended next steps. Use conditional language — this is a preliminary evaluation, not a legal determination.

Example output (paraphrased): The model used appropriate conditional language (“potential exists for recovery against…”) and outlined evidence preservation steps clearly.

Watch for: AI will sometimes suggest contacting attorneys or filing suits as routine next steps. Subrogation referrals are internal decisions — keep the note focused on evidence and potential, not action items above your authority level.

Prompt 7: Denial Letter Draft

When to use: When you need to draft a coverage denial letter for supervisor review. This is a high-stakes output — never send AI-drafted denial letters without attorney review and supervisor sign-off.

The prompt:

Draft a coverage denial letter for an insurance claim.

Carrier name: [Name]
Claimant name: [Name]
Claim number: [Number]
Date of loss: [Date]
Reason for denial: [e.g., "Damage is consistent with long-term seepage, excluded under policy section [X] as gradual damage not caused by a sudden and accidental event"]
Policy section cited: [e.g., "Section I, Exclusion 8(a): 'We do not cover loss resulting from continuous or repeated seepage or leakage of water that occurs over a period of 14 or more days'"]

Draft a denial letter in professional insurance claims language. Include: (1) description of loss, (2) our investigation findings, (3) applicable policy language, (4) basis for denial, (5) right to appeal language. Tone: professional and factual. Do not be adversarial.

Example output (paraphrased): The model produced a well-structured letter with all five required elements. The right-to-appeal language was appropriately standard.

Watch for: This is the one output where I always have a supervisor or in-house counsel review before anything goes to the claimant. AI denial letters look correct but may cite policy language incorrectly or omit required state-mandated language. In some states, denial letter requirements are statutory.

Prompt 8: Estimate Review Summary

When to use: When you’re reviewing a contractor’s estimate or a public adjuster’s estimate and need to document your analysis.

The prompt:

Write an estimate review summary for a claim file.

Estimate submitted by: [Contractor / Public adjuster / Restoration company]
Estimate amount: $[amount]
Our Xactimate estimate: $[amount]
Variance: $[amount] ([X]%)
Key line items in dispute: [e.g., "Code upgrade allowance: contractor at $4,200 vs. our $1,100; overhead and profit: contractor at 20%/10% vs. our 10%/10%"]

Write a 200-word estimate review summary in claims language. State the variance, identify key disputed items, explain the basis for our position on each disputed item, and recommend next steps (negotiation, appraisal, or payment of our estimate).

Example output (paraphrased): The model correctly structured the comparison and identified each disputed item with a defensible position. The recommended next steps were appropriate and specific.

Watch for: Overhead and profit disputes are heavily fact-specific and often litigated. Do not let AI summarize your O&P position without reviewing Xactimate’s published O&P guidelines and your carrier’s handling instructions.

Tips for Customizing These Prompts

Add your carrier’s language standards. If your carrier uses specific phrases like “sudden and accidental” as a technical term, add a note: “Use the phrase ‘sudden and accidental event’ when describing covered water losses. Do not use ‘unexpected.’” Consistency matters when claims get audited.

Include your name and claim number in every prompt. It sounds obvious, but having the AI generate report language with the correct claim number embedded means less editing later.

Set word limits. Add “Keep the total response under 300 words” to avoid AI padding your reports with unnecessary background. Adjuster reports should be thorough, not long.

Tell the model what not to do. I’ve found negative instructions (“Do not speculate about coverage,” “Do not include caveats about consulting an attorney”) are as important as positive ones. AI defaults to CYA language that has no place in a field report.

Common Prompt Mistakes in Insurance Work

Pasting policy language and asking AI to interpret it. AI will confidently misread policy exclusions, especially manuscript policies or endorsements. Use AI for formatting and structure; do your own coverage analysis.

Using AI for sworn statement questions. If you’re preparing for an Examination Under Oath or a recorded statement, AI can help you organize topics but should not script the questions. EUO strategy depends on claim-specific facts and legal considerations that AI cannot assess.

Skipping the review because the output looks right. A generated causation paragraph that says “abrupt pipe failure” when your notes say “slow leak” is a coverage analysis error waiting to happen. The output will look polished and professional right up until a plaintiff’s attorney uses it against you.

Not saving your prompts. Once you find a prompt that produces consistently good output for your claim type, save it. A library of 8-10 tested prompts for your most common claim types is worth more than experimenting with each new claim.