The short answer
Underwriters can use AI prompts to summarize submission documents, draft risk narratives, compare policy forms, and check facts against underwriting guidelines. The key is to never paste un-redacted personally identifiable information (PII) into a generic AI tool. Effective prompts provide context, specify the desired output format, and use placeholders for sensitive data to remain compliant.
AI prompts are a powerful way for insurance underwriters and other professionals to speed up routine work, from triaging new business submissions to documenting a final decision. But for a role that handles vast amounts of sensitive client data, the biggest risk isn’t the AI getting a fact wrong—it’s the underwriter violating data privacy laws. ZEKAI is an independent AI-tools directory, and in this guide, we’ll provide more than 25 copy-paste prompts specifically for underwriters, with a focus on how to get the benefits of AI without the compliance risks.
This guide is built for the licensed professional whose decisions have legal weight. We’ll cover the specific regulations that govern your use of AI and provide practical, safe prompts you can use today. For a broader view of how AI is impacting all roles in the industry, see our guide to AI for insurance claims, risk, and analytics.
The One Rule: Never Paste Client PII into a Public AI
Before we get to the prompts, we must establish the ground rule for using any public AI tool like ChatGPT, Claude, or Gemini for insurance work. You cannot, under any circumstances, paste nonpublic personal information (NPI) or personally identifiable information (PII) into the free, consumer versions of these tools.
(GLBA) Source: fdic.gov
The Gramm-Leach-Bliley Act (GLBA) legally requires financial institutions—including insurance companies—to explain their information-sharing practices and safeguard sensitive data. Pasting a client’s name, address, date of birth, policy number, or any financial or medical information into a public AI tool can constitute a data breach and a violation of the GLBA’s Safeguards Rule.
By default, consumer-grade AI tools may use your prompts to train their models. Even if you use settings to opt out of training, your conversations may be retained for 30 days for abuse monitoring and can be reviewed by humans.
The solution is rigorous redaction. Treat AI as an untrusted third party. Before you paste any document or data, remove:
- Names of individuals or companies
- Addresses, phone numbers, email addresses
- Policy numbers, account numbers, driver’s license numbers
- Dates of birth, Social Security numbers
- Specific financial figures (use representative numbers if needed, e.g., “$5M limit” instead of “$5,127,500”)
- Any information that could uniquely identify an individual or entity.
All prompts in this guide are designed with this redaction-first principle in mind.
Submission Intake & Triage Prompts
The first bottleneck in underwriting is the flood of incoming submissions, each with its own structure and attachments. These prompts help you quickly extract key facts and determine if a submission is worth a deeper look.
I am a commercial property underwriter. I have received a broker submission for a new business account. The attached documents include:
1. ACORD 125 & 126 forms
2. 5 years of currently valued loss runs
3. A statement of values for 15 locations
Please act as an underwriting assistant. Review the following [redacted] text from these documents and extract the key data points into a structured summary. I need to know:
- The proposed policy effective date.
- The applicant's industry (NAICS code if available).
- Total insured value (TIV) across all locations.
- Number of locations with a value over $[threshold, e.g., 10M].
- Total number of open claims and total incurred on closed claims in the past 3 years.
- Any large losses over $[threshold, e.g., 100k].
Here is the redacted text:
[Paste redacted text from ACORDs, SOV, and loss runs]
I am reviewing a new business submission for a [line of business, e.g., Cyber Liability] policy. Based on our standard underwriting requirements listed below, please review the provided [redacted] submission documents and identify any missing information or documents that I need to request from the broker.
Our requirements:
- Completed and signed application
- 3 years of financial statements
- Details of current cybersecurity controls (MFA, EDR, etc.)
- Incident response plan
- List of all third-party vendors with access to sensitive data
Provided documents:
[Paste redacted text or list of document titles]
Act as a senior underwriter. My firm's risk appetite for [line of business, e.g., General Liability for contractors] is attached. Please review the following summary of a [redacted] new business submission and identify any characteristics that fall outside our stated appetite or require special consideration.
My firm's appetite:
- No more than 25% of work is residential.
- No work on buildings over 5 stories.
- No wrap-up / OCIP projects.
- Must have been in business for at least 3 years.
Submission summary:
[Paste a redacted summary of the applicant's operations, e.g., "Applicant is a framing contractor, 5 years in business. 40% of recent work has been on residential multi-family projects up to 7 stories. Has one OCIP project on the books for next year."]
Please draft a professional and courteous email to an insurance broker declining to quote a submission. The reason for declination is that the applicant's operations fall outside our underwriting appetite for [specific reason, e.g., 'residential construction exposure']. The email should be polite, thank the broker for the opportunity, and express interest in seeing other business from them that fits our appetite. Do not include any specific details about the applicant.
I have an inbox full of emails from brokers. Please categorize the following email subjects into three buckets:
1. **New Business Submissions:** Requiring full review.
2. **Endorsement/Change Requests:** Requiring processing on an existing policy.
3. **General Inquiries:** Questions that need a response.
Email subjects:
[Paste a list of email subject lines, e.g., "FW: New GL/Auto Submission - ABC Construction", "Bind Order Request - Policy #12345", "Question on XYZ Corp Renewal", "Can you write a car wash?"]
Risk Assessment & Guideline-Mapping Prompts
Once a submission is triaged, the real work of underwriting begins. These prompts help you analyze the risk, compare it to your company’s guidelines, and start building the narrative for your decision.
I am analyzing a renewal for a client who currently has our [Policy Form A] and is considering moving to [Policy Form B]. Please act as a coverage expert. I will provide the key coverage and exclusion sections from both [redacted] policy forms. Your task is to create a table that compares the key differences in coverage for:
- [Coverage part 1, e.g., Definition of "Insured"]
- [Coverage part 2, e.g., Water Damage Exclusion]
- [Coverage part 3, e.g., Notice of Claim provision]
Highlight which form provides broader coverage for the insured in each case.
[Paste redacted text from both policy forms]
I have 5 years of [redacted] loss runs for a workers' compensation renewal. The data is in a messy format. Please extract and summarize this information into a clear table with columns for: Year, Total Incurred, Total Paid, Number of Claims, and a brief, one-sentence description of any claim over $[threshold, e.g., 50,000]. After the table, provide a 2-3 sentence summary of the overall loss trend.
[Paste redacted loss run data]
Act as a compliance analyst. I am underwriting a [type of risk, e.g., a restaurant with liquor service]. My company's underwriting guidelines are provided below. Please review the [redacted] applicant information and create a checklist that shows whether the applicant meets, fails, or needs more information for each guideline.
Guidelines:
- Must have a fire suppression system over cooking surfaces.
- Staff must have alcohol service training.
- No history of liquor license violations.
- Security plan required if open past midnight.
Applicant Info:
[Paste redacted applicant details, e.g., "Applicant is a 150-seat restaurant, open until 2 AM on weekends. Has a wet chemical fire suppression system. All servers complete TIPS training. No security plan mentioned in the application."]
I am preparing for an underwriting call with the broker for a complex [type of risk, e.g., manufacturing] account. Based on the initial [redacted] submission information below, generate a list of 10 specific, open-ended questions I should ask to better understand the risk, their safety procedures, and their quality control.
Submission Info:
[Paste redacted summary, e.g., "Applicant manufactures and imports widgets from an overseas supplier and distributes them to big-box retailers. They have a product recall plan but the details are vague."]
Using the structured data points provided below, write a 2-paragraph risk summary for an underwriting file. The summary should be objective and concise, suitable for a manager's review. It should cover the nature of the operations, the positive attributes of the risk, and the key concerns or exposures that need to be addressed in pricing or terms.
Structured Data:
- **Line of Business:** [e.g., Commercial Auto]
- **Applicant:** [e.g., Regional plumbing contractor, 15 years in business]
- **Fleet Size:** [e.g., 25 vehicles]
- **Radius of Operations:** [e.g., 100 miles]
- **Driver Vetting:** [e.g., MVR checks annually]
- **Safety Program:** [e.g., Telematics in all vehicles, monthly driver training]
- **Loss History:** [e.g., One large liability loss 3 years ago, otherwise clean]
- **Concern:** [e.g., Vehicle usage includes emergency calls, raising nighttime driving risk]
Loss Control & Communication Prompts
Good underwriting isn’t just about saying yes or no; it’s about helping good risks become better. These prompts help you communicate with brokers and insureds about risk improvement.
A [type of risk, e.g., retail store] I insure has experienced several slip-and-fall claims in the past year. Based on this pattern, generate a list of 5 practical loss control recommendations I can suggest to the insured to help prevent future incidents. Frame them as helpful advice.
Please draft a letter to a broker outlining the subjectivities that must be met before coverage can be bound for a [line of business] policy. The subjectivities are:
1. Receipt of a signed wind mitigation report.
2. Confirmation that the fire alarm system is centrally monitored.
3. A copy of the contract with their main supplier.
The tone should be clear, professional, and collaborative.
I need to explain to a broker why the "Fungi or Bacteria" exclusion is being added to their client's property policy. Please draft a simple, clear explanation (about 100 words) that I can use in an email. The explanation should avoid overly technical jargon and focus on the fact that this is a standard exclusion for un-mitigated water damage and mold, which is a common industry practice for this type of risk.
I have a 20-page property inspection report for a risk I am underwriting. I will paste the [redacted] 'Recommendations' and 'Key Deficiencies' sections. Please summarize the findings into a few bullet points, separating critical life-safety recommendations from standard maintenance suggestions.
[Paste redacted text from the inspection report]
I need to have a difficult conversation with a broker about a non-renewal due to deteriorating loss experience. Let's role-play. I am the underwriter. You are the broker. I will start by explaining the situation. Your goal is to push back, ask for another chance, and question my data. I need you to help me practice responding calmly and sticking to the facts of the case. Start by responding to this: "Hi [Broker Name], I'm calling about the renewal for XYZ Corp. I'm afraid I have some bad news. We won't be able to offer a renewal quote this year.
Renewal & Portfolio Review Prompts
Underwriting isn’t a one-time event. These prompts help with the ongoing management of a book of business.
I am preparing for the renewal of a large [line of business] account. The policy expires in 90 days. Please create a checklist of information I need to gather and actions I need to take. Include items like:
- Requesting updated exposure information from the broker.
- Analyzing the current term's loss experience.
- Checking for any changes in the insured's operations.
- Reviewing the current pricing against new rate-and-form filings.
I have a [redacted] spreadsheet of my current book of business, with columns for [Policy Number, NAICS Code, Premium, Incurred Losses]. Please act as a data analyst and describe the overall characteristics of my portfolio. I want to know:
- What are the top 3 industries by premium?
- What is the overall loss ratio for the book?
- Are there any industry segments with a loss ratio significantly higher than the average?
[Paste redacted spreadsheet data]
Based on the following [redacted] portfolio data, identify the top 5 policies that are the strongest candidates for a significant rate increase at renewal. Your criteria should be a combination of a high loss ratio over the past 3 years and a current premium that appears low relative to their exposure. List the policy IDs and a one-sentence justification for each.
[Paste redacted data with columns for Policy ID, 3-Year Loss Ratio, and Current Premium/Exposure]
I have the expiring terms for a policy and the proposed renewal terms. Please create a table comparing the two, showing the changes in:
- Premium
- Limits
- Deductibles
- Key endorsements being added or removed
Expiring: [e.g., $100k premium, $1M limit, $5k deductible]
Renewal: [e.g., $120k premium, $1M limit, $10k deductible, new water damage exclusion added]
Please draft a short paragraph for a broker to share with their client, explaining why their [e.g., Cyber] insurance premium is increasing at renewal even though they have had no claims. The explanation should reference general market conditions, such as the increased frequency and severity of ransomware attacks across the industry and the rising cost of reinsurance for all carriers.
Compliance & File Documentation Prompts
The final step is documenting your work to withstand audits and regulatory scrutiny.
2026, over half of all U.S. states had adopted the NAIC’s Model Bulletin on the Use of AI Systems by Insurers or similar guidance. Source: waterstreetcompany.com
This means regulators expect you to have a documented process for how you use AI. These prompts help create that documentation.
I have decided to [decline/quote/bind] a policy for [redacted applicant ID]. Based on the following summary of facts, please generate a concise, objective, and compliant entry for the underwriting file that documents the basis for my decision. The entry should reference the specific underwriting guidelines that were met or not met.
Facts:
- Risk: [e.g., New venture restaurant]
- Guideline 1: Requires 3 years of management experience. Applicant has 10 years. (Met)
- Guideline 2: Requires fire suppression system. Applicant does not have one. (Not Met)
- Guideline 3: Location must not be in a flood zone. Location is in Flood Zone A. (Not Met)
- Decision: Decline.
Act as a compliance officer. Colorado's SB 21-169 prohibits unfair discrimination based on protected classes in insurance. I am about to decline a risk. Please review my proposed declination reason below and flag if it could be perceived as a proxy for a protected class or otherwise be considered unfairly discriminatory.
Reason for Declination: "The applicant's credit score is below our minimum threshold.
I used a generative AI model to help summarize a large submission file. I need to document this for my underwriting file to comply with the NAIC AI Model Bulletin. Please draft a short note for the file that states:
- What tool was used (e.g., "Internal instance of a large language model").
- What task it was used for (e.g., "To summarize 50 pages of redacted loss runs").
- That no PII was entered into the model.
- That I, the licensed underwriter, personally reviewed the AI-generated summary for accuracy and made the final underwriting decision based on my own judgment and the verified source documents.
I need to write a file note that another underwriter might have to read years from now. Please translate the following internal underwriting shorthand into clear, unambiguous language.
Shorthand: "Sent to UW mgr for auth bc of TIV. Will need reins. support. Subj to survey."
Clear Language: [Your output should be something like: "Referred to the underwriting manager for approval due to the Total Insured Value exceeding my binding authority. This risk will require facultative reinsurance support. Binding is subject to a satisfactory loss control survey."]
Imagine I am an insurer who has received a request from a state regulator about our use of AI in underwriting. Draft a template for a response that confirms our commitment to fair and compliant practices. The response should mention:
- Our adherence to the NAIC AI Model Bulletin.
- Our established AI governance framework.
- The use of human oversight and judgment for all final underwriting decisions.
- Our policy of not using AI to make binding decisions without a human in the loop.
Where AI Stops and Licensed Judgment Starts
No AI prompt can replace the core responsibility of a licensed professional. An AI can summarize, draft, and analyze, but it cannot exercise binding authority. “Binding authority” is the legal power granted by an insurer to an underwriter to accept a risk on the insurer’s behalf. That power, and the legal accountability that comes with it, remains with the human underwriter.
When you use AI, you are still the one signing off on the decision. Tools like the AI platforms from SAS Insurance AI Suite or Gradient AI can provide sophisticated risk scores and insights, but they are decision-support tools. The final call—and the liability for it—is yours. Using AI responsibly means using it to make your human judgment faster and better-informed, not to replace it.
As AI becomes more integrated into insurance workflows, understanding both its capabilities and its legal limitations is the most critical skill for the modern insurance professional. To learn more about how different roles are adapting, visit our AI in insurance profession hub.
Can AI legally make an underwriting decision?
No. As of September 2026, an AI cannot legally bind an insurance policy. Binding authority rests with a licensed human underwriter who is accountable for the decision. AI systems are used for decision support—to score risk and summarize data—but the final, legally binding decision must be made by a person.
Is it safe to paste applicant data into ChatGPT?
No. Pasting nonpublic personal information (NPI) or personally identifiable information (PII) into the free, public version of ChatGPT or other general AI tools can violate privacy laws like the GLBA and break your company’s data security policies. Data must be fully redacted before being used in such tools.
Will AI replace insurance underwriters?
No, but AI is changing the underwriter’s job. AI automates routine data entry and analysis, allowing underwriters to focus on more complex risks, broker relationships, and portfolio strategy. Professionals who learn to use AI as a tool to enhance their judgment will become more valuable, not obsolete.
What is the NAIC AI Model Bulletin?
The NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers is a set of guidelines for the responsible use of AI. Adopted by many states, it expects insurers to have a formal governance program, manage risk, and ensure AI use is fair and transparent, holding the insurer ultimately responsible for all decisions.
How do you document AI use for a compliance audit?
For an audit, you should document what tool you used, what you used it for (e.g., summarizing documents), and critically, that no sensitive client data was input. Your file note should affirm that you, the licensed underwriter, validated the AI’s output and made the final decision based on your professional judgment.
Where to go next
Three routes, picked for what you just read.
Sources (24)
- | URL |
- |—|
- | https://www.fdic.gov/consumers/consumer/news/glba.html |
- | https://www.waterstreetcompany.com/blog/what-the-naic-model-bulletin-means-for-insurance-ai |
- | https://content.naic.org/cipr-topics/artificial-intelligence-ai |
- | https://dora.colorado.gov/news-releases/sb21-169-protecting-consumers-from-unfair-discrimination-in-insurance-practices |
- | https://www.sonomos.com/blog/free-vs-paid-chatgpt-privacy-2026 |
- | https://www.spellbook.legal/blog/is-chatgpt-private |
- | https://www.v7labs.ai/blog/generative-ai-in-insurance |
- | https://www.insuranceopedia.com/definition/binding-authority/2/6703 |
- | https://www.legis.nd.gov/cencode/t26-1c26-3.pdf |
- | https://www.pibit.ai/blog/what-is-delegated-underwriting-authority-dua |
- | https://www.willkie.com/offices/washington/insights/2023/12/the-grammleachbliley-act-glba-privacy-rule |
- | https://www.egnyte.com/guides/governance/glba-compliance |
- | https://www.transcend.io/blog/glba-compliance |
- | https://www.pii-tools.com/a-beginners-guide-to-glba-compliance/ |
- | https://www.sonant.ai/blog/insurance-binder-guide |
- | https://www.rsmus.com/insights/industries/financial-services/insurance/colorado-new-york-take-the-lead-on-regulating-insurers-use-of-a.html |
- | https://www.credo.ai/company-blog/colorado-sb21-169-8-things-you-need-to-know-about-colorados-new-ai-insurance-regulation |
- | https://leg.colorado.gov/bills/sb21-169 |
- | https://www.plantemoran.com/our-insights/articles/2026/how-the-naic-ai-model-bulletin-is-evolving |
- | https://www.hollandknight.com/en/insights/publications/2025/05/the-implications-and-scope-of-the-naic-model-bulletin-on-the-use-of-ai-by-insurers |
- | https://www.clydeco.com/en/insights/2024/08/naic-model-bulletin-use-of-artificial-intelligen |
- | https://www.proton.me/blog/is-chatgpt-private |
See Zekai first in Google
The weekly AI briefing for your profession
One weekly email: the AI changes that actually affect your profession — tools, deals, and what to do about them.



